Many of today's most important applications rely on high-performance manipulation of matrices, tensors, and general arrays of data. The neural networks at the heart of AI language models are a prime example. The AI software stack relies on kernels (e.g., for matrix-matrix product) that are specifically designed to deliver throughput close to the hardware's theoretical peak. However, writing kernels that can deliver maximum performance is extremely challenging. In this talk, I will discuss some of the work we have done to simplify the task of authoring such kernels. This includes new abstractions to allow for algebraic manipulation of array layouts and new programming models for more easily expressing the kernels themselves.
Michael Garland has been a researcher at NVIDIA since 2006. He joined the company as one of the founding members of NVIDIA Research and is currently Senior Director of Programming Systems Research. He leads a research group focused on developing technologies that will help programmers take advantage of modern high-performance machines. Their work spans the software stack with a particular focus on parallel algorithms, programming languages, compilers and runtime systems, and low-level hardware/software interfaces. He and his team have both published numerous academic articles and contributed to many software packages in broad use by CUDA developers.
Drawing on her research in pricing, revenue management, and market design, Dr. Wang explores how leaders can make better pricing and selling decisions when market data is limited, noisy, or constantly changing. Her work focuses on simple, practical strategies that remain effective under uncertainty, helping organizations improve revenue outcomes and design more reliable market mechanisms. This session will offer a timely look at how rigorous analytics can guide smarter decisions in imperfect real-world conditions.
Featuring Dr. Shixin Wang, Tennenbaum Early Career Professor and Assistant Professor in the H. Milton Stewart School of Industrial and Systems Engineering (ISyE) at Georgia Institute of Technology. Prior to joining Georgia Tech, Dr. Wang was an Assistant Professor in the Department of Decisions, Operations and Technology at The Chinese University of Hong Kong (CUHK) Business School from 2021 to 2025. Her research interests lie in developing simple and robust pricing policies in revenue management, and designing sparse and reliable networks in supply chain and service systems.
Classical complexity theory characterizes the difficulty of optimization problems through global worst-case parameters---Lipschitz constants, smoothness, dimension---and derives bounds that are tight over the entire problem class. Yet algorithms routinely outperform these predictions in practice, prompting the question of which other structural properties may determine algorithm efficiency. In this talk, I will present a line of research showing that local structural properties of optimization problems can reveal tractability hidden by worst-case analysis. I will discuss two interconnected threads. The first develops a new complexity framework based on local subgradient variation, which captures when optimization is substantially easier than global bounds suggest. A striking consequence is that parallelization can provably accelerate convex optimization for broad problem classes---including those underlying classical lower bounds---overturning long-standing conventional wisdom. The second thread applies a similar "closer look" philosophy to robust learning: I will discuss how local geometric structure enables polynomial-time algorithms with best-possible error guarantees for learning generalized linear and single-index models under adversarial noise and distributional shifts. I will close with future directions toward a broader theory of optimization that explains efficiency beyond the worst case.
]]>Optimization Meets Participation: Iterative School Zone Generation with LLMs
In U.S. public school systems, geographic boundaries play a central role in shaping students’ assignments and access to opportunity. For example, the San Francisco Unified School District (SFUSD) recently adopted multi-school zones with controlled choice to jointly promote diversity and proximity to assigned schools. Designing such zones is both computationally and socially complex: algorithmic approaches are required to balance competing objectives at scale, yet stakeholders are typically asked to articulate their preferences upfront, before seeing feasible zone maps, limiting their ability to meaningfully influence outcomes. We propose a stakeholder-in-the-loop framework for joint preference elicitation and zone design. Our approach iterates between using optimization to generate zones and collecting participatory feedback as stakeholders react to zones. To enable broad participation, we use large language models (LLMs) to translate between natural language stakeholder input and optimization constraints. To support real-time iteration, we develop faster computational methods for the multi-school zoning problem, using both mathematical programming and sampling-based approaches. Our framework produces zones with substantially improved diversity and proximity metrics relative to existing benchmarks, while also generating individual-level preference representations that can be aggregated using standard social choice methods. Our approach has been used to support preliminary discussions about zone boundaries in SFUSD and are generalizable to other redistricting and participatory planning contexts.
Irene Lo is an assistant professor in the department of Management Science & Engineering at Stanford University. Her research sits at the intersection of operations research, computer science theory, and economic theory. She designs markets and allocation systems that improve both efficiency and equity, with applications in education, the environment, and the developing world. She leads a Stanford Impact Lab on Equitable Access to Education, co-launched the ACM Conference series on Equity and Access in Algorithms, Mechanisms, and Optimization (EAAMO), and is a William T. Grant Scholar.
]]>Distributionally Robust Newsvendor on a Metric
We consider a fundamental generalization of the classical newsvendor problem where the seller needs to decide on the inventory of a product jointly for multiple locations on a metric as well as a fulfillment policy to satisfy the uncertain demand that arises sequentially over time after the inventory decisions have been made. To address the distributional-ambiguity, we consider a distributionally robust setting where the decision-maker only knows the mean and variance of the demand, and the goal is to make inventory and fulfillment decisions to minimize the worst-case expected inventory and fulfillment cost (where the expectation is taken over the worst case choice of distribution with given mean and variance).
We present a significant generalization of the classical result of Scarf (1958) and give a policy with strong theoretical guarantees as well as good practical performance while maintaining the simplicity and interpretability of the solution in Scarf (1958). In particular, our policy first identifies a hierarchical clustering of the locations, and assigns a "virtual-underage cost" for each cluster. Our inventory solution ensures that for each cluster, the total inventory in the cluster is at least as large as the inventory level suggested by Scarf's solution for the virtual-underage cost if the cluster was a single point. We present a worst-case performance guarantee for our policy and also demonstrate that the policy performs well in practice. To the best of our knowledge, this is the first algorithm with provable performance guarantees. (This is joint work with Ayoub Foussoul)
Vineet Goyal is a Professor in the Industrial Engineering and Operations Research Department at Columbia University where he joined in 2010. He received his Bachelor's degree in Computer Science from Indian Institute of Technology, Delhi in 2003 and his Ph.D. in Algorithms, Combinatorics and Optimization (ACO) from Carnegie Mellon University in 2008. Before coming to Columbia, he spent two years as a Postdoctoral Associate at the Operations Research Center at MIT. He is interested in the design of efficient and robust data-driven algorithms for large scale dynamic optimization problems with applications in revenue management and healthcare problems. His research has been continually supported by grants from NSF and industry including NSF CAREER Award in 2014 and faculty research awards from Google, IBM, Adobe and Amazon.
]]>
We present a significant generalization of the classical result of Scarf (1958) and give a policy with strong theoretical guarantees as well as good practical performance while maintaining the simplicity and interpretability of the solution in Scarf (1958). In particular, our policy first identifies a hierarchical clustering of the locations, and assigns a "virtual-underage cost" for each cluster. Our inventory solution ensures that for each cluster, the total inventory in the cluster is at least as large as the inventory level suggested by Scarf's solution for the virtual-underage cost if the cluster was a single point. We present a worst-case performance guarantee for our policy and also demonstrate that the policy performs well in practice. To the best of our knowledge, this is the first algorithm with provable performance guarantees. (This is joint work with Ayoub Foussoul)
]]>The digital twin, a virtual representation of a physical system or process, integrates information obtained from sensor data, physics models, as well as operational and inspection/maintenance/repair history of the system. As more and more data becomes available, the resulting updated model becomes increasingly accurate in predicting future behavior of the system, and can potentially be used to support several objectives, such as sustainment, mission planning, and operational maneuvers. This presentation will present recent research in digital twin methodologies to support all three objectives, based on several types of computations: current state diagnosis, model updating, future state prognosis, and decision-making. All these computations are affected by uncertainty regarding system properties, operational parameters, usage and environment, as well as uncertainties in data and the prediction models. Therefore the presentation will address decision-making under uncertainty, and the incorporation of modern uncertainty quantification techniques, considering both aleatory and epistemic uncertainty sources. Scaling up the probabilistic digital twin methodology to support real-time decision-making is a challenge, and several strategies that combine recent advances in sensing, computing, data fusion, and machine learning to enable the scale-up will be discussed. Several use cases related to power grid, aircraft, marine vessels, and additive manufacturing will be presented.
Professor Sankaran Mahadevan (Vanderbilt University, Nashville, TN) has more than thirty-five years of research and teaching experience in uncertainty quantification, risk and reliability analysis, machine learning, structural health diagnosis and prognosis, and decision-making under uncertainty. He has applied these methods to a variety of structures, materials and systems in civil, mechanical and aerospace engineering. His research has been extensively funded by NSF, NASA, DOE, DOD, FAA, NIST, as well as GM, Chrysler, GE, Union Pacific, and Mitsubishi, and he has co-authored two textbooks and 350 peer-reviewed journal papers. During the past two decades, he has been at the forefront of academic research on uncertainty quantification and digital twin methodologies.
Professor Mahadevan has served as President of the ASCE Engineering Mechanics Institute, and as chair of several technical committees and prominent conferences in ASCE, ASME, and AIAA. He is currently serving as the Chair of the ASME VVUQ 50 Subcommittee on Advanced Manufacturing. He is a Distinguished Member of ASCE, and Fellow of AIAA, Engineering Mechanics Institute (ASCE), and PHM Society. His awards include ASCE’s Alfredo Ang award for risk analysis and management of civil infrastructure, and the IASSAR Distinguished Research award.
]]>Featuring Jon Duke, a physician-scientist with a 25-year career spanning clinical practice, academic medicine, and advanced research in health informatics and analytics. Dr. Duke joined Georgia Tech Research Institute (GTRI) in 2016 and has served as director of GTRI’s Health Emerging and Advanced Technologies Division since its inception in 2019. Over the course of his career, Dr. Duke has been awarded over $70M in external funding as a principal investigator and his academic publications have been cited over 6,000 times. Dr. Duke's work is enabling organizations to better use electronic health data to improve the well-being of patients and populations. His areas of research include decision support, data standards and interoperability, natural language processing, and artificial intelligence with applications spanning medication safety, public health, cost transparency, and clinical research.
]]>
My team’s work on AI for Social Impact (AI4SI) has spanned two decades, focusing on optimizing limited resources in critical areas like public health, conservation, and public safety. I will present field results from India, where the deployment of restless and collaborative multi-armed bandit (RMAB) algorithms achieved significant improvements in the world’s two largest mobile maternal health programs. I will also highlight ongoing work on network-based HIV prevention in South Africa, modeled as a branching bandit problem. These projects, along with other initiatives across Africa and Asia, expose a critical "deployment bottleneck" that spans the entire machine learning pipeline. This bottleneck consists of three key hurdles: the observational scarcity gap (data), the policy synthesis gap (learning and modeling), and the human-AI alignment gap (deployment).
This talk investigates how Generative AI can address this AI4SI deployment bottleneck through the strategic use of LLM Agents and diffusion models. I will demonstrate how LLM Agents address the alignment gap by integrating expert guidance into algorithmic planning, ensuring resource optimization strategies reflect real-world priorities. Furthermore, I will show how diffusion models mitigate the scarcity and synthesis gaps by generating synthetic social networks and facilitating Transfer RL to utilize data across domains. I will conclude by discussing this path toward a more scalable, human-aligned future for AI for Social Impact.
Bio:
Milind Tambe is the Gordon McKay Professor of Computer Science at Harvard University; concurrently, he is Principal Scientist and Director for “AI for Social Good” at Google Research. Prof. Tambe and his team have developed innovative AI and multi-agent reasoning systems that have been successfully deployed to deliver real-world impact in public health (e.g., maternal and child health), public safety, and wildlife conservation. He is the recipient of the AAAI Award for Artificial Intelligence for the Benefit of Humanity, the AAAI Feigenbaum Prize, the IJCAI John McCarthy Award, the AAAI Robert S. Engelmore Memorial Lecture Award, and the AAMAS ACM/SIGAI Autonomous Agents Research Award. He is a fellow of AAAI and ACM. His contributions in Operations Research and public safety have also been recognized with the INFORMS Wagner Prize for excellence in Operations Research practice, Military Operations Research Society Rist Prize, the Columbus Fellowship Foundation Homeland security award, and commendations and certificates of appreciation from the US Coast Guard, the Federal Air Marshals Service, and airport police at the city of Los Angeles.
This talk investigates how Generative AI can address this AI4SI deployment bottleneck through the strategic use of LLM Agents and diffusion models. I will demonstrate how LLM Agents address the alignment gap by integrating expert guidance into algorithmic planning, ensuring resource optimization strategies reflect real-world priorities. Furthermore, I will show how diffusion models mitigate the scarcity and synthesis gaps by generating synthetic social networks and facilitating Transfer RL to utilize data across domains. I will conclude by discussing this path toward a more scalable, human-aligned future for AI for Social Impact.
]]>
Many manufacturing companies have experienced costly recalls and product failures because uncertainties in design, testing, and manufacturing processes were not adequately quantified. These failures have led to fatal accidents, billions of dollars in lost revenue, and even the collapse of major manufacturing firms. In response, industries such as aerospace, automotive, semiconductor, and medical devices have increasingly adopted Uncertainty Quantification (UQ)—a multidisciplinary framework drawing from statistics, applied mathematics, and engineering—to better design, test, and manufacture products under uncertainty.
This talk provides an overview of Uncertainty Quantification, explains why it has become indispensable in modern engineering, and introduces key design of experiment and predictive model methods for rigorously quantifying uncertainty in complex systems.
Peter Chien is a Professor of Statistics and Industrial & Systems Engineering at the University of Wisconsin–Madison and a Fellow of the American Statistical Association. He is the recipient of a National Science Foundation CAREER Award and an IBM Faculty Award. His research has been widely adopted by Fortune 500 companies across industries including aerospace, automotive, semiconductors, electronics, chemical, battery and life sciences.
This talk provides an overview of Uncertainty Quantification, explains why it has become indispensable in modern engineering, and introduces key design of experiment and predictive model methods for rigorously quantifying uncertainty in complex systems.
]]>XR for the Sky: Enhancing UAV and UAM Operations
Georgia Tech, College of Engineering : DR-CSE-AE, Labs: ASDL and CCG (Aerospace Systems Design Laboratory and Contextual Computing Group)
We actively encourage cross-departmental participation and welcome engagement from both internal academic units and external industry partners. Our goal is to foster a robust community of practice around XR at Georgia Tech.
To join our mailing list and RSVP your attendance for these seminar series, please sign up here: RSVP HERE. Lunch will be served for those who RSVP.
]]>Veridical Data Science for Healthcare in the Age of AI
Georgia Tech Exhibition Hall
Kirkwood Room
Monday, March 30, 2026
3:30-4:30PM
Reception to follow at ISyE Main Atrium
Data science underpins modern AI and many advances in healthcare, yet human judgment permeates every stage of the data science life cycle. These judgment calls introduce hidden uncertainties that go well beyond sampling variability and drive many of the risks associated with AI.
We introduce veridical data science, grounded in three fundamental principles—Predictability, Computability, and Stability (PCS)—to make such uncertainties explicit and assessable and to aggregate reality-checked algorithms for better results. The PCS framework unifies and extends best practices in statistics and machine learning and is illustrated through healthcare applications, including identifying genetic drivers of heart disease, reducing cost of prostate cancer detection, improving uncertainty quantification beyond standard conformal prediction, and proposing, Green Shielding, a new user-centric framework for safeguarding users of AI.
Dr. Bin Yu is CDSS Chancellor's Distinguished Professor in Statistics, EECS, Center for Computational Biology, and Senior Advisor at the Simons Institute for the Theory of Computing, all at UC Berkeley. Her research focuses on the practice and theory of statistical machine learning, veridical data science, responsible and safe AI, and solving interdisciplinary data problems in neuroscience, genomics, and precision medicine. She and her team have developed algorithms such as iterative random forests (iRF), stability-driven NMF, adaptive wavelet distillation (AWD), Contextual Decomposition for Transformers (CD-T), SPEX and ProxySPEX for interpreting deep learning models, especially for compositional interpretability.
She is a member of the National Academy of Sciences and of the American Academy of Arts and Sciences. She was a Guggenheim Fellow, President of Institute of Mathematical Statistics (IMS), and delivered the Tukey Lecture of the Bernoulli Society, the Breiman Lecture at NeurIPS, the IMS Rietz Lecture, and the Wald Memorial Lectures (the highest honor of IMS), and Distinguished Achievement Award and Lecture (formerly Fisher Lecture) of COPSS (Committee of Presidents of Statistical Societies). She holds an Honorary Doctorate from The University of Lausanne. She is on the Editorial Board of Proceedings of National Academy of Science (PNAS) and a co-editor of the Harvard Data Science Review (HDSR).
]]>We strongly encourage you to attend to seek full-time employment, internships, and projects (rather than waiting until the end of the semester).
If you are interested in hosting a table for our upcoming session, please let us know after reviewing the below information within our website. Early registration closes August 1st!
Visit https://www.scl.gatech.edu/outreach/supplychainday.
]]>Robot-Independent Visual-Tactile XR Teleoperation for Multi-Humanoid Cooperation
George W. Woodruff School of Mechanical Engineering (Georgia Tech College of Engineering), Symbiotic and Augmented Intelligence Laboratory (SAIL)
We actively encourage cross-departmental participation and welcome engagement from both internal academic units and external industry partners. Our goal is to foster a robust community of practice around XR at Georgia Tech.
To join our mailing list and RSVP your attendance for these seminar series, please sign up here: RSVP HERE. Lunch will be served for those who RSVP.
]]>LitForager: Exploring Multimodal Literature Foraging Strategies in Immersive Sensemaking
IVI Lab (Immersive Visualization & Interaction Lab), School of Interactive Computing (Georgia Tech College of Computing)
We actively encourage cross-departmental participation and welcome engagement from both internal academic units and external industry partners. Our goal is to foster a robust community of practice around XR at Georgia Tech.
We actively encourage cross-departmental participation and welcome engagement from both internal academic units and external industry partners. Our goal is to foster a robust community of practice around XR at Georgia Tech.
To join our mailing list and RSVP your attendance for these seminar series, please sign up here: RSVP HERE. Lunch will be served for those who RSVP.
]]>AHEAD of Time: Toward Robots That Behave Like Human Companions
George W. Woodruff School of Mechanical Engineering (Georgia Tech College of Engineering), Symbiotic and Augmented Intelligence Laboratory (SAIL)
We actively encourage cross-departmental participation and welcome engagement from both internal academic units and external industry partners. Our goal is to foster a robust community of practice around XR at Georgia Tech.
To join our mailing list and RSVP your attendance for these seminar series, please sign up here: RSVP HERE. Lunch will be served for those who RSVP.
]]>Motion, Heart Rate, and Haptics: Using XR devices for Human Monitoring and Regulation
School of Interactive Computing (Georgia Tech College of Computing), Georgia Tech Sonification Lab
We actively encourage cross-departmental participation and welcome engagement from both internal academic units and external industry partners. Our goal is to foster a robust community of practice around XR at Georgia Tech.
To join our mailing list and RSVP your attendance for these seminar series, please sign up here: RSVP HERE. Lunch will be served for those who RSVP.
]]>Simulating the Future: Experience Prototyping for Emerging Technologies via Multimodal XR
School of Psychology (Georgia Tech College of Sciences), Georgia Tech Sonification Lab
We actively encourage cross-departmental participation and welcome engagement from both internal academic units and external industry partners. Our goal is to foster a robust community of practice around XR at Georgia Tech.
To join our mailing list and RSVP your attendance for these seminar series, please sign up here: RSVP HERE. Lunch will be served for those who RSVP.
]]>Human-in-the-Loop and XR for Context-Rich Industrial Decision-Making
We actively encourage cross-departmental participation and welcome engagement from both internal academic units and external industry partners. Our goal is to foster a robust community of practice around XR at Georgia Tech.
To join our mailing list and RSVP your attendance for these seminar series, please sign up here: RSVP HERE. Lunch will be served for those who RSVP.
Solving the Mysteries of Place Cells and Grid Cells by Representation Learning
The 2014 Nobel Prize in Physiology or Medicine recognized the discovery of place cells and grid cells in the mammalian brain. Each place cell fires at a single specific location, whereas each grid cell fires at multiple locations forming a hexagonal grid pattern. Yet the computational principles underlying these phenomena have remained mysterious. We show both emerge from representation learning through geometric optimization. Grid cells learn embeddings that preserve local distances through conformal isometry, forming a coordinate system. We prove hexagonal patterns are optimal: hexagonal flat tori uniquely minimize deviation from local distance preservation by distributing curvature isotropically through six-fold symmetry. Building upon this coordinate system, place cells learn embeddings that preserve spatial adjacency relations defined by transition kernels of heat diffusion with reflecting boundary conditions, thereby forming a cognitive map. Specifically, inner products between embeddings reconstruct transition probabilities, causing localized firing patterns to emerge automatically from non-negative matrix factorization constraints. This reveals how the brain solves navigation by transforming spatial reasoning into optimization on learned geometric representations.
Ying Nian Wu is a professor in the Department of Statistics and Data Science at UCLA. He earned his A.M. and Ph.D. in statistics from Harvard University in 1994 and 1996, respectively. From 1997 to 1999, he served as an assistant professor in the Department of Statistics at the University of Michigan before joining UCLA in 1999. He became a full professor in 2006, and he was an Amazon Scholar 2020-2025. Wu’s research spans generative AI, computer vision, computational neuroscience, and bioinformatics.
]]>This course provides a deep dive into the ways in which artificial intelligence (AI) optimizes supply chain efficiency. Participants will explore generative AI fundamentals, prompt engineering, and practical applications such as automated inventory systems, predictive maintenance, and route optimization. The course also covers ethical AI use, good and bad use of generative AI (GenAI), and rapidly emerging use cases. By the end, professionals will be skilled in applying AI to enhance supply chain processes and drive success in their organizations.
This course targets supply chain managers, data analysts, logistics professionals, procurement specialists, and business leaders aiming to harness GenAI for enhanced supply chain operations. It is ideal for those interested in GenAI-driven efficiency, strategic insights, and navigation of GenAI's role in transforming supply chain processes.
Making AI Impactful in Healthcare
There is increasing evidence that Machine Learning and Artificial intelligence algorithms can be used to enhance clinical care. In this talk, I address two critical aspects that can significantly improve the impact of such algorithms in healthcare practices: (1) moving beyond associations and creating algorithms capable of causal reasoning under ambiguity, and (2) a human-algorithm “centaur” model of care and decision-making, in which the power of human intuition is combined with the outstanding capabilities of algorithms. I describe our latest research on these subjects at the Public Impact Analytics Science Lab (PIAS-Lab) at Harvard, and discuss findings based on our various collaborations with the Mayo Clinic, Mass General Hospital, and some other public and private organizations.
Soroush Saghafian (Wikipedia) is an Associate Professor at Harvard University and is the founder and director of Harvard’s Public Impact Analytics Science Lab (PIAS-Lab). He also serves as a core faculty or a faculty affiliate for (a) Harvard Data Science Initiative, (b) Harvard Mossavar-Rahmani Center for Business and Government, (c) Harvard Center for Health Decision Science, (d) Harvard Ph.D. Program in Health Policy, (e) Harvard Belfer Center for Science and International Affairs, (f) Harvard Center for Public Leadership, and (g) Harvard Ariadne Labs (a pioneer lab in health systems innovation), and holds appointments at Massachusetts General Hospital (MGH), Beth Israel Deaconess Medical Center (BIDMC), and Mayo Clinic. He is an expert in healthcare AI, analytics, and operations management, and has collaborated with a variety of hospitals.
Dr. Saghafian's research has appeared numerous times in the news including in top national and international media outlets, and has been recognized through various awards, including the INFORMS MSOM Young Scholar Prize for “outstanding contributions to scholarship in operations management,” INFORMS MSOM Responsible Research Award (second place) for “contributing knowledge that may have implications for making the world a better place,” the Inaugural INFORMS Mehrotra Research Excellence Award “for significant contributions to the practice of health applications through operations research and management science modeling and methodologies,” INFORMS Computing Society Harvey Greenberg Award (honorable mention) “for research excellence in the field of computation and operations research applications, especially those in emerging application fields,” INFORMS Pierskalla Award “for the best research paper in healthcare,” INFORMS Franz Edelman Award (semi-finalist) “for achievement in advanced analytics, operations research, and management science,” and POMS College of Healthcare Best Paper Award. His forthcoming book with Cambridge University Press, “Insight-Driven Problem Solving: Analytics Science to Improve the World,” has been endorsed by top academic and industry figures [Full CV here].
]]>There is increasing evidence that Machine Learning and Artificial intelligence algorithms can be used to enhance clinical care. In this talk, I address two critical aspects that can significantly improve the impact of such algorithms in healthcare practices: (1) moving beyond associations and creating algorithms capable of causal reasoning under ambiguity, and (2) a human-algorithm “centaur” model of care and decision-making, in which the power of human intuition is combined with the outstanding capabilities of algorithms. I describe our latest research on these subjects at the Public Impact Analytics Science Lab (PIAS-Lab) at Harvard, and discuss findings based on our various collaborations with the Mayo Clinic, Mass General Hospital, and some other public and private organizations.
]]>Thursday, March 5, 2026 | 12-1pm ET
Electric vehicles are reshaping freight networks, charging demand is growing faster than infrastructure can keep up, and companies are trying to understand how electrification will change the cost and design of their supply chains. Georgia Tech Assistant Professor, Dr. Constance Crozier’s research tackles these questions with models that capture grid constraints, charging behavior, transportation patterns, and the economic tradeoffs behind electrified logistics. In this session, she will walk through the forces that will define the future of freight electrification.
Thursday, April 2, 2026 | 12-1pm ET
AI initiatives often struggle not because of model sophistication, but because the underlying data supply chain has not been mapped, synchronized, or aligned with operational decision-making. Just as a physical supply chain cannot function without visibility, coordination, and flow from source to destination, AI systems cannot create value when data sources are siloed, inconsistently governed, or disconnected from the “last mile” of decision execution. Even well designed models stall when the upstream data inputs are unreliable or the downstream decision processes are unclear.
This talk reframes AI implementation through the lens of the hidden data supply chain that propels it - from data sourcing and acquisition, to transformation, integration, governance, and delivery into decision environments. We will explore the continuous loop between data engineering and AI development, showing how model requirements should shape data architecture from the outset, and why data pipelines must be engineered as critical infrastructure rather than reactive fixes. Participants will leave with a practical framework for mapping their data supply chain, identifying bottlenecks and failure points in the data-to-decision flow, and building resilient data architectures that support reliable, explainable, and production ready AI systems.
Featuring Rosemarie Santa González, Ph.D., Research Scientist in the Institute for Robotics and Intelligent Machines (IRIM) and the NSF AI‑CARING Institute at Georgia Tech, and an instructor in the SCL professional education program.
January 23, 2026
10:30 AM – 4:00 PM
ISyE Main Building
Agenda:
10:30 AM – Noon (ISyE Main Atrium):
XR Symposium: Invited Talks - Short talks and discussion on XR research and applications across campus
Noon – 1:00 PM (ISyE Main Atrium):
Lunch & Networking: Informal networking in the ISyE Main Atrium
1:00 – 4:00 PM (ISyE Main 115)
XR Makerspace Open House & Demos: Hands-on demonstrations, guided tours, and conversations with XR Makerspace staff and researchers
Invited Speakers:
Bruce Walker, School of Psychology & School of Interactive Computing, Georgia Tech
Talk Title: Research With, About, and In Virtual Reality: A Sample of Sonification Lab XR R&D Projects
Bio: Bruce N. Walker is Professor of Psychology and Interactive Computing at Georgia Tech. His Sonification Lab studies multimodal interfaces, sonification and auditory displays, and human-technology interaction in complex tasks, futuristic technologies, and VR/AR/XR. Dr. Walker is founding Director of VRlandia, a GT shared collaboratory for VR/AR/XR research and development. He is also Director of the Center for Human-AI-Robot Interaction (CHART). He teaches HCI, Sensation & Perception, and Assistive Technology; and has consulted for NASA, state and federal governments, the military, and private companies.
Frederick Benaben, H. Milton Stewart School of Industrial & Systems Engineering, Georgia Tech
Talk Title: A Framework for Immersive Technologies in Engineering: An Illustrated Perspective
Bio: Frederick Benaben is Tenured Full Professor at Institut Mines Telecom in France (Industrial Engineering Center) and Visiting Professor at the Georgia Institute of Technology. His research focuses on the use of Artificial Intelligence and Immersive Technologies for decision support and the management of complex situations in uncertain environments. He is the head of the “Digital Systems for Crisis Management and Security” research team at IMT, directs the IOMEGA VR Lab, leads the POD (Physics of Decision) and HITeC (Hybrid Immersive Teaching Campus) initiatives, and co-directs the international SIReN Lab with Georgia Tech.
Shreyas Kousik, George W. Woodruff School of Mechanical Engineering, Georgia Tech
Talk Title: Do Dangerous Things Safely in XR Humanoid Robot Teleoperation
Bio: Shreyas Kousik is an assistant professor in the Mechanical Engineering at Georgia Tech. He previously held postdoctoral appointments at Stanford University and in the NAV Lab with Prof. Grace Gao. He earned his Ph.D. in Mechanical Engineering from the University of Michigan, advised by Prof. Ram Vasudevan in the ROAHM Lab. His research focuses on guaranteeing safety in autonomous robotic systems through collision avoidance, with an emphasis on modeling uncertainty in perception and estimation for practical planning and control on real robots.
Wayne Li, Colleges of Design and Engineering, Georgia Tech
Talk Title: Using XR for Automotive Interior Development (Hyundai Project)
Bio: Wayne K. Li is the James L. Oliver Professor, holding a joint position between the Colleges of Design and Engineering at Georgia Institute of Technology. He leads joint teaching initiatives and advances interdisciplinary collaboration between mechanical engineering and industrial design through classes and the Innovation and Design Collaborative (IDC) and Design Bloc. Li's research areas include ethnographic research, multidisciplinary online education, and human-machine interaction in transportation design. His career spans industry and academia. Li has led innovation and market expansion for Pottery Barn seasonal home products, taught in Stanford University's design program, led interface development at Volkswagen of America's Electronics Research Laboratory, and developed corporate brand and vehicle differentiation strategies at Ford Motor Company. He has also worked as a product designer and mechanical engineer at IDEO Product Development.
Meryem Yilmaz Soylu, Center for 21st Century Universities, Georgia Tech
Talk Title: Designing Human-Centered XR for Learning, Reflection, and Skill Development
Bio: Meryem Yilmaz Soylu is a Research Scientist at Georgia Tech’s Center for 21st Century Universities (C21U) within the College of Lifetime Learning, where her work focuses on human-centered learning design, immersive technologies, and AI-supported educational systems. Her research examines how XR and AI can support learning experiences and engagement, leadership, and durable skills development across online and hybrid learning environments. Her work combines mixed-methods research and learner experience (UX) design to inform the design and evaluation of real-world educational systems.
]]>The free course will address how to prevent injury and illness due to workplace exposures in warehouses.
Each session to be held at the below location:
Professional Administration Resource Building
Georgia Tech Savannah Campus
210 Technology Circle
Savannah, GA 31408
We will have the capability to provide on-site training. Training materials and course content will be offered at no cost to interested facilities. oshainfo.gatech.edu
]]>
The free course will address how to prevent injury and illness due to workplace exposures in warehouses.
Each session to be held at the below location:
Professional Administration Resource Building
Georgia Tech Savannah Campus
210 Technology Circle
Savannah, GA 31408
We will have the capability to provide on-site training. Training materials and course content will be offered at no cost to interested facilities. oshainfo.gatech.edu
]]>
Leading with Purpose and Innovation: Manhattan Associates' Journey
ISyE Main
Thursday, February 26, 2026
3:30-4:30PM
Reception to follow.
Over the past 25 years, Manhattan Associates has played a pivotal role in transforming global supply chains, enabling many of the world’s largest retailers, brands, pharmaceutical distributors, and logistics providers to operate faster, smarter, and more reliably.
In this lecture, Eddie Capel, chairman of the board at Manhattan Associates, will reflect on Manhattan’s journey and evolution during a period of continuous technological disruption, from the early days of the internet, through the shift to cloud-native platforms, and into today’s AI-driven era. He will discuss how being an early adopter of emerging technologies, combined with disciplined engineering thinking and foundation, allowed Manhattan to build and continuously reinvent its solutions, and sustain industry leadership.
The session highlights how strategic technology choices, long-term R&D investment, and a strong culture enabled Manhattan to deliver significant gains in supply chain efficiency and productivity. Eddie will also share leadership lessons for the next generation, emphasizing curiosity, adaptability, and the importance of embracing new technologies early to shape, rather than react to, the future of the overall industry.
Eddie Capel serves as the executive chairman of the board at Manhattan Associates. He previously served as executive vice-chairman for three months from February 2025 until end of 2025. Before that, beginning in 2013, Mr. Capel led the company as its president and CEO, driving innovation and growth. Prior to that, he served as executive vice president and chief operating officer. With more than 30 years of experience in supply chain strategy and operations, Mr. Capel has played a pivotal role in shaping industry-leading solutions.
Before joining Manhattan Associates in June 2000, Mr. Capel held key leadership positions at Real Time Solutions (RTS), where he served as chief operations officer and vice president of operations. In these roles, he led teams that supported the supply chain strategies of major companies, including Walmart, Amazon.com, and J.C. Penney. He also served as director of operations at Unarco Automation, an Industrial Automation/Robotics systems integrator. Earlier in his career, he worked as a project manager and system designer for ABB Robotics in the United Kingdom.
]]>PM2.5, which contributes to 90% of the total particle mass emitted during these events. Because these particles are small enough to enter the bloodstream, the ability to "see," "track," and "predict" the spread of smoke has become an urgent task with significant societal implications.This threat is especially acute for vulnerable populations, including children, pregnant women, older adults, and outdoor workers, as well as those with preexisting cardiovascular disease or from low socioeconomic status groups. Developing a robust tracking capability is critical for generating accurate, real-time air quality predictions during active wildfire episodes, allowing these individuals to take timely and effective mitigation actions.
Beyond immediate safety, advanced modeling provides a vital dataset for scientific investigations into the long-term health impacts of wildfires. By leveraging the generative nature of modern AI, we can efficiently simulate multiple smoke spread scenarios under various conditions. This facilitates long-term planning for essential services, such as managing hospital capacity during fire events, directing traffic, and informing insurance policy.
This talk will present preliminary results on the modeling and prediction of wildfire smoke using a combination of remote-sensing and computer simulation data. We will also explore future research opportunities in this rapidly evolving field of environmental health and data science.
Featuring Xiao Liu, PhD, David M. McKenney Family Associate Professor, H. Milton Stewart School of Industrial and Systems Engineering, Georgia Tech Center for Health and Humanitarian Systems. Dr. Liu's research focuses on developing data-driven methods for scientific and engineering applications. His work has been published in leading Industrial Engineering and Statistics journals, including JASA, and AOAS. He has served as the president of the Data Analytics & Information Systems division of IISE and as Program Chair for the 2025 IISE Annual Conference & Expo. Before returning to academia, he worked as a research staff member at the IBM Thomas J. Watson Research Center.
Offered online via Zoom. Please register to attend.
]]>Microtransit offers opportunities to enhance urban mobility by combining the reliability of public transit and the flexibility of ride-sharing. This paper optimizes the design and operations of a deviated fixed-route microtransit system that relies on reference lines but can deviate on demand in response to passenger requests. We formulate a Microtransit Network Design (MiND) model via two-stage stochastic integer optimization, with a first-stage network design and service scheduling structure and a second-stage vehicle routing structure. We derive a tight second-stage relaxation using a subpath-based representation of microtransit operations in a load-expanded network. We develop a double-decomposition algorithm combining Benders decomposition and subpath-based column generation. We prove that the algorithm maintains a valid optimality gap and converges to an optimal solution in a finite number of iterations. Results obtained with real-world data from Manhattan show that the methodology scales to large and otherwise-intractable instances, with up to 10-100 candidate lines and hundreds of stops. Comparisons with transit and ride-sharing suggest that microtransit can provide win-win outcomes toward efficient mobility (high demand coverage, low costs, high level of service), equitable mobility (broad geographic reach) and sustainable mobility (limited environmental footprint).
]]>Frontiers and Applications at the Interface of Discrete Optimization and Interpretable Machine Learning
Modern machine learning models achieve remarkable predictive accuracy and can capture complex interactions, but they are often difficult to interpret and may fail to reveal useful relationships in the data. This lack of interpretability also limits their use in high-stakes applications such as healthcare, where predictions must be auditable for trust and safety. Using tree ensembles (e.g., gradient boosting or random forests) as a motivating example, we propose a novel optimization-based framework for extracting interpretable rule-based models at the post-training stage. We formulate rule extraction as a large-scale discrete optimization problem that balances predictive accuracy with considerations such as model compactness, stability, and transparency. To address these problems, we develop specialized algorithms that scale beyond the capabilities of off-the-shelf optimization software. Using mental telehealth treatment data from our industry collaborators at SilverCloud Health, we demonstrate how these methods enable practitioners to extract meaningful insights from complex datasets and predictive models.
Brian Liu is a fifth-year Ph.D. candidate in Operations Research at MIT, advised by Professor Rahul Mazumder. His research lies at the intersection of discrete optimization, statistics, and computer science, with a focus on developing efficient and interpretable machine learning algorithms. His work is motivated by real-world applications in domains such as healthcare and medicine and has received multiple Best Student Paper Awards from INFORMS (Data Mining; Quality, Statistics, and Reliability) and the American Statistical Association (Statistical Computing; Nonparametric Statistics).
]]>Modern machine learning models achieve remarkable predictive accuracy and can capture complex interactions, but they are often difficult to interpret and may fail to reveal useful relationships in the data. This lack of interpretability also limits their use in high-stakes applications such as healthcare, where predictions must be auditable for trust and safety. Using tree ensembles (e.g., gradient boosting or random forests) as a motivating example, we propose a novel optimization-based framework for extracting interpretable rule-based models at the post-training stage. We formulate rule extraction as a large-scale discrete optimization problem that balances predictive accuracy with considerations such as model compactness, stability, and transparency. To address these problems, we develop specialized algorithms that scale beyond the capabilities of off-the-shelf optimization software. Using mental telehealth treatment data from our industry collaborators at SilverCloud Health, we demonstrate how these methods enable practitioners to extract meaningful insights from complex datasets and predictive models.
]]>Don’t miss this opportunity to explore cutting-edge AI applications in Medical and Health Care Systems.
https://sites.gatech.edu/ai-mhcs/
]]>Don’t miss this opportunity to explore cutting-edge AI applications in Medical and Health Care Systems.
]]>Featuring Lauren N. Steimle, PhD, Assistant Professor, H. Milton Stewart School of Industrial and Systems Engineering, Georgia Tech Center for Health and Humanitarian Systems. Dr. Steimle's research involves the creation of industrial engineering and operations research (IE/OR) methodologies to answer decision-making problems arising in public health and medicine. Her areas of research include medical decision-making, regionalized systems of healthcare delivery, and infectious disease prevention and control. Her work is motivated by addressing fundamental problems arising from challenges in maternal health, poliovirus, and chronic disease management, and COVID-19.
Offered online via Zoom. Please register to attend.
]]>Bridging Machine Learning and Optimization for Human-Centered AI
From healthcare delivery to resilient power grid management, predictive and prescriptive analytics tools have the potential to improve decision-making for some of today’s most pressing problems, yet their impact is often limited by the technical barriers required to access these tools and to interpret and trust their results. This talk will explore how the synthesis of machine learning and optimization can lower these barriers to advance human-centered artificial intelligence (AI). The first part of the talk will demonstrate how generative AI can broaden access to optimization tools through an interactive decision-support framework, developed in collaboration with Microsoft Outlook, that leverages large language models to translate natural-language user requests into underlying constraint programming models. The second part of the talk will focus on trust, showing how optimization can identify regions where machine learning models make fixed predictions that preclude individuals from changing their outcomes, such as a loan applicant who can never be approved regardless of their actions. We will conclude by outlining broader opportunities for integrating AI and optimization, moving toward a future in which advanced analytics tools are as accessible and trustworthy for managers at a local food bank as they are for applied scientists at Amazon.
Connor Lawless is a Postdoctoral Fellow at the Stanford Institute for Human-Centered Artificial Intelligence advised by Ellen Vitercik and Madeleine Udell. His research blends tools from optimization, machine learning, and human-computer interaction to make advanced analytics tools more accessible and trustworthy. He received his PhD in Operations Research from Cornell University where he was advised by Oktay Gunluk, and previously spent time at Microsoft Research, IBM Research, and the Royal Bank of Canada.
]]>From healthcare delivery to resilient power grid management, predictive and prescriptive analytics tools have the potential to improve decision-making for some of today’s most pressing problems, yet their impact is often limited by the technical barriers required to access these tools and to interpret and trust their results. This talk will explore how the synthesis of machine learning and optimization can lower these barriers to advance human-centered artificial intelligence (AI). The first part of the talk will demonstrate how generative AI can broaden access to optimization tools through an interactive decision-support framework, developed in collaboration with Microsoft Outlook, that leverages large language models to translate natural-language user requests into underlying constraint programming models. The second part of the talk will focus on trust, showing how optimization can identify regions where machine learning models make fixed predictions that preclude individuals from changing their outcomes, such as a loan applicant who can never be approved regardless of their actions. We will conclude by outlining broader opportunities for integrating AI and optimization, moving toward a future in which advanced analytics tools are as accessible and trustworthy for managers at a local food bank as they are for applied scientists at Amazon.
]]>Thursday, February 5, 2026 | 12-1pm ET
Emergencies push supply chains into conditions they were never designed for. Demand surges, resources become scarce, and leaders must make fast decisions with limited information. Dr. Mathieu Dahan will share insights from his research on how service systems and workforce models perform under stress, and how flexible capacity and smarter allocation policies can improve response outcomes. The session is designed for leaders responsible for continuity planning, public-private coordination, and operational risk, offering a clear view of what more resilient response systems can look like.
Every fall and spring semester, during the IISE Career Fair, companies across the nation come to Georgia Tech to recruit some of the nation’s top talent from our Bachelor’s and Master’s programs. Our students are recruited for a variety of roles and perform well past expectations in all positions. We hope that you will join us this semester and meet some of the country’s brightest students. This event is open to all majors, and recruiting for roles in Data Analytics, Supply Chain, Consulting, Operations, Finance, Business, Tech & more.
Recruit from the #1-Ranked Industrial Engineering Program in the nation at the IISE Career Fair.
Explore internships and full-time opportunities with companies recruiting across Industrial & Systems Engineering and related fields. For more info & resume submission, visit https://www.gtiise.org/career-fair-info
Secure Your Spot Early for the Spring 2026 Career Fair! For questions, email iise@gatech.edu.
]]>Uncertainty quantification for black-box models with conditional guarantees
A central problem in uncertainty quantification is designing methods that are both distribution-free and individualized to the test sample at hand. Prior work has shown that it is impossible to achieve finite-sample conditional validity without modelling assumptions. Thus, canonical methods in the conformal inference literature typically only issue marginal guarantees over a random draw of the test covariates. In this talk, I will outline a framework that bridges this gap by recasting the conditional objective as a set of robustness criteria under covariate shifts. By modifying the target class of covariate shifts, I will define a spectrum of problems that range between marginal and exact instance-wise validity and give methods that provide precise guarantees in between these extremes. This framework has broad applications and I will show how it can be used to construct prediction sets around the outputs of black-box regression models and filter out false information from the responses of large language models. This talk is based on joint work with John Cherian and Emmanuel Candès.
Isaac Gibbs is a postdoctoral researcher at the University of California, Berkeley, where he is advised by Ryan Tibshirani. He received his Ph.D. in Statistics from Stanford University, advised by Emmanuel Candès, and his B.Sc. in Math and Computer Science from McGill University. His research focuses on topics related to predictive inference, distribution-free uncertainty quantification, online learning, and high-dimensional statistics.
]]>A central problem in uncertainty quantification is designing methods that are both distribution-free and individualized to the test sample at hand. Prior work has shown that it is impossible to achieve finite-sample conditional validity without modelling assumptions. Thus, canonical methods in the conformal inference literature typically only issue marginal guarantees over a random draw of the test covariates. In this talk, I will outline a framework that bridges this gap by recasting the conditional objective as a set of robustness criteria under covariate shifts. By modifying the target class of covariate shifts, I will define a spectrum of problems that range between marginal and exact instance-wise validity and give methods that provide precise guarantees in between these extremes. This framework has broad applications and I will show how it can be used to construct prediction sets around the outputs of black-box regression models and filter out false information from the responses of large language models. This talk is based on joint work with John Cherian and Emmanuel Candès.
]]>GPU-Accelerated Linear Programming and Beyond
The rapid progress in GPU computing has revolutionized many fields, yet its potential in mathematical programming, such as linear programming (LP), has only recently begun to be realized. This talk aims to provide an overview of recent advancements in GPU-based first-order methods for LP, with a particular focus on the design and development of cuPDLPx. The extensions to GPU-based optimization beyond LP, including convex quadratic programming and semidefinite programming, will also be discussed.
Jinwen Yang is a final-year Ph.D. student at the University of Chicago, advised by Professor Haihao Lu. His research interests are in optimization, with a particular focus on optimization algorithms tailored to modern hardware (like GPUs) and intended for practical applications. He obtained B.S. in Mathematics and Applied Mathematics from Fudan University.
]]>The rapid progress in GPU computing has revolutionized many fields, yet its potential in mathematical programming, such as linear programming (LP), has only recently begun to be realized. This talk aims to provide an overview of recent advancements in GPU-based first-order methods for LP, with a particular focus on the design and development of cuPDLPx. The extensions to GPU-based optimization beyond LP, including convex quadratic programming and semidefinite programming, will also be discussed.
]]>Don’t miss this opportunity to explore cutting-edge AI applications in Medical and Health Care Systems.
]]>Don’t miss this opportunity to explore cutting-edge AI applications in Medical and Health Care Systems.
]]>Subscribe to our newsletter: CHHS Newsletter
Learn more about CHHS: Center for Health and Humanitarian Systems
]]>
This course provides a deep dive into the ways in which artificial intelligence (AI) optimizes supply chain efficiency. Participants will explore generative AI fundamentals, prompt engineering, and practical applications such as automated inventory systems, predictive maintenance, and route optimization. The course also covers ethical AI use, good and bad use of generative AI (GenAI), and rapidly emerging use cases. By the end, professionals will be skilled in applying AI to enhance supply chain processes and drive success in their organizations.
This course targets supply chain managers, data analysts, logistics professionals, procurement specialists, and business leaders aiming to harness GenAI for enhanced supply chain operations. It is ideal for those interested in GenAI-driven efficiency, strategic insights, and navigation of GenAI's role in transforming supply chain processes.
Many Supply Chain decisions are concerned with the timely and efficient procurement, allocation, and distribution of resources (e.g. funds, supplies, volunteers, money, employees) through a supply chain network. This course will explore methodologies for “medium term” decision making including procurement and inventory policies, strategies for distribution and allocation of limited resources, and supply chain design.
This course is designed for representatives from governmental or non-governmental organizations, private corporations, military, and foundations, including but not limited to senior executives overseeing administrative and operational functions of an organization, logistics and supply chain managers, program managers, directors of field operations, directors of emergency/disaster preparedness and response, and public health professionals.
This course is the second in a 3-part virtually synchronous professional education program. Register and pay for all three required Health and Humanitarian Supply Chain Management Certificate courses and receive a discount of $400 off per course. Enter coupon code SCL-HHS at checkout with the Georgia Tech Professional Education website..
Additionally, there are scholarships available for the certificate program. Apply at https://chhs.gatech.edu/course-scholarships by the noted deadline.
Questions? Reach out to chhs@gatech.edu!
]]>This course provides a deep dive into the ways in which artificial intelligence (AI) optimizes supply chain efficiency. Participants will explore generative AI fundamentals, prompt engineering, and practical applications such as automated inventory systems, predictive maintenance, and route optimization. The course also covers ethical AI use, good and bad use of generative AI (GenAI), and rapidly emerging use cases. By the end, professionals will be skilled in applying AI to enhance supply chain processes and drive success in their organizations.
This course targets supply chain managers, data analysts, logistics professionals, procurement specialists, and business leaders aiming to harness GenAI for enhanced supply chain operations. It is ideal for those interested in GenAI-driven efficiency, strategic insights, and navigation of GenAI's role in transforming supply chain processes.
This landmark edition will begin in Albi, where the pre-conference on June 5–6 will bring together PhD students, researchers, and industry professionals for interactive training sessions, including the “My Thesis in 180 Seconds” challenge, and immersive learning experiences using cutting-edge Virtual Reality games. The event will then move to Bordeaux from June 8–10 for the main conference, themed “Supply Chain and Logistics Intelligence in the Era of the Physical Internet: Bridging High-Tech and Low-Tech Solutions.”
Participants will enjoy a vibrant program featuring scientific and industrial sessions, inspiring keynotes, dynamic roundtables, exhibitions, company visits, and collaborative workshops designed to foster innovation and shape the future of logistics and supply chain networks.
And, because logistics is global, participants will be from all over the world including researchers, industrial and international institution members, local authorities and standardization committees.
About the Physical Internet Initiative
The Physical Internet Initiative aims at transforming the way physical objects are moved, stored, realized, supplied and used, pursuing global logistics efficiency and sustainability. Originating from Professor Benoit Montreuil in 2006, this ground breaking vision, revolutionizing current paradigms, has stirred great interest from scientific, industrial as well as governmental communities.
Open to Georgia Tech students from all majors — ISyE, Business, CS, ECE, and beyond — to encourage broad participation and interdisciplinary exchange.
The event is designed to bring together Georgia Tech students who are exploring AI in supply chain, logistics, and related fields. The goal is to share ideas, compare approaches, and build connections across majors and programs.
If you’ve used AI in any way—whether for a class, research, or a side project—we’d love to hear from you. Presentations are brief and informal (3–5 minutes), and slides are optional. Your insights can spark great conversations!
We will review registrations to ensure a good balance of presenters and listeners for an interactive session. If you’re presenting, you’ll share a short talk. If you’re attending as a listener, we expect you to actively engage—ask questions, share feedback, and join the discussion during networking. Your input helps make this session valuable for everyone!
Click here to register Online to Attend
Thursday, January 8, 2026 | 12-1pm ET
From disruptive technologies to shifting global dynamics, the supply chain landscape is evolving faster than ever. Whether you're navigating nearshoring strategies, exploring AI-driven logistics, or simply trying to keep pace with change, this session will help you spot what's coming and act before the competition does.
A Statistical Framework for Benchmarking Quantum Computers
]]>
The last two decades have witnessed quantum computing technologies increasingly move from theoretical proposals to functioning experimental platforms, reaching major milestones such as the demonstration of beyond-classical computational tasks. The central challenge in the sustained development of large-scale quantum computers is the presence of hardware errors, which must be identified and quantified before they can be mitigated. In this talk, I will develop a statistical perspective on this problem of benchmarking quantum devices, using an experimental platform known as random circuit sampling. Data arising from this experiment can be described through a high-dimensional discrete latent variable model parametrized by hardware-level error rates. We develop computationally efficient estimators for these error rates, which incorporate side information about the model via simulations from a reference quantum computer. These estimators achieve the information-theoretic estimation limits for this problem, implying that reliable estimation is possible even for large-scale quantum devices that evade classical computational abilities. We apply our methods to benchmark a recent state-of-the-art quantum processor, obtaining a detailed report with hundreds of error rates which were largely unavailable from past studies. I will conclude by placing these results in the broader context of my interdisciplinary work in the physical sciences, and by discussing some of my other research interests in nonparametric statistics and statistical optimal transport.
]]>Heterogeneous Treatment Effects in Panel Data: Applications to the Healthy Incentives Program
In this talk, we will discuss work motivated by studying the Healthy Incentives Program (HIP), a food-subsidy program. Our goal is to quantify how adding new vendors affects program utilization using observational panel data. In particular, the effects may be heterogeneous, and the timing of the interventions may be highly irregular. This is an instance of a core problem in causal inference: estimating heterogeneous treatment effects using panel data with general treatment patterns.
To address this problem, we introduce the Panel Clustering Estimator (PaCE). PaCE partitions observations into clusters with similar treatment effects using a regression tree and leverages the low-rank structure of the panel data to estimate the average treatment effect within each cluster. Our theoretical results identify conditions on the treatment patterns under which the treatment effects are recoverable, and we establish convergence guarantees under those conditions. Computational experiments show that PaCE achieves higher accuracy than existing approaches while remaining interpretable. Applying PaCE to HIP data, we identify the heterogeneous impacts of vendor additions on HIP utilization across Massachusetts ZIP codes and uncover key demographic and contextual factors driving these differences. Our findings provide valuable insights for future budget planning and for identifying which ZIP codes to target with vendor additions.
Emily Zhang is a fifth-year PhD candidate at the MIT Operations Research Center, advised by Professors Retsef Levi and Georgia Perakis. Her research addresses critical challenges in food systems, focusing on reducing food waste and improving equitable access to healthy food. Her work spans modeling, optimization, causal inference, inventory management, and data-driven operations, and has been conducted in collaboration with the Massachusetts Department of Transitional Assistance and nonprofit organizations such as Met Council. Prior to her PhD, she earned dual B.S. degrees in Computer Science and Mathematics from MIT.
]]>In this talk, we will discuss work motivated by studying the Healthy Incentives Program (HIP), a food-subsidy program. Our goal is to quantify how adding new vendors affects program utilization using observational panel data. In particular, the effects may be heterogeneous, and the timing of the interventions may be highly irregular. This is an instance of a core problem in causal inference: estimating heterogeneous treatment effects using panel data with general treatment patterns.
To address this problem, we introduce the Panel Clustering Estimator (PaCE). PaCE partitions observations into clusters with similar treatment effects using a regression tree and leverages the low-rank structure of the panel data to estimate the average treatment effect within each cluster. Our theoretical results identify conditions on the treatment patterns under which the treatment effects are recoverable, and we establish convergence guarantees under those conditions. Computational experiments show that PaCE achieves higher accuracy than existing approaches while remaining interpretable. Applying PaCE to HIP data, we identify the heterogeneous impacts of vendor additions on HIP utilization across Massachusetts ZIP codes and uncover key demographic and contextual factors driving these differences. Our findings provide valuable insights for future budget planning and for identifying which ZIP codes to target with vendor additions.
]]>Seeing the Forest for the Trees
Abstract:
Recent papers in the graph machine learning literature have introduced a number of approaches for hyperbolic representation learning. The asserted benefits are improved performance on a variety of graph tasks, node classification and link prediction included. Claims have also been made about the geometric suitability of particular hierarchical graph datasets to representation in hyperbolic space. Despite these claims, our work makes a surprising discovery: when simple Euclidean models with comparable numbers of parameters are properly trained in the same environment, in most cases, they perform as well, if not better, than all introduced hyperbolic graph representation learning models, even on graph datasets previously claimed to be the most hyperbolic as measured by Gromov delta-hyperbolicity (i.e., perfect trees).
This observation gives rise to a simple question: how can this be? We answer this question by taking a careful look at the field of hyperbolic graph representation learning as it stands today, and find that a number of results do not diligently present baselines, make faulty modelling assumptions when constructing algorithms, and use misleading metrics to quantify geometry of graph datasets. We take a closer look at each of these three problems, elucidate the issues, perform an analysis of methods, and introduce a parametric family of benchmark datasets to ascertain the applicability of (hyperbolic) graph neural networks.
Unfortunately, these problems are not specific to hyperbolic graph neural nets but are indicative of more fundamental problems in graph machine learning more generally. We show, surprisingly, that node features are oftentimes more-than-sufficient for many common graph benchmarks, breaking this critical assumption. When comparing against a well-tuned feature-only MLP baseline on seven of the most commonly used graph learning datasets, one gains little benefit from using graph structure on five datasets. We posit that these datasets do not benefit considerably from graph learning because the features themselves already contain enough graph information to obviate or substantially reduce the need for the graph.
Bio:
Anna Gilbert received her S.B. degree from the University of Chicago and a Ph.D. from Princeton University, both in Mathematics. In 1997, she was a postdoctoral fellow at Yale University and AT&T Labs-Research. From 1998 to 2004, she was a member of technical staff at AT&T Labs-Research in Florham Park, NJ. From 2004 to 2020, she was with the Department of Mathematics (with a secondary appointment in Electrical and Computer Engineering) at the University of Michigan, where she eventually became the Herman H. Goldstine Collegiate Professor. In 2020, she moved to Yale University as the John C. Malone Professor of Mathematics and Professor of Statistics & Data Science. In 2023, she left the Mathematics Department and is now in Statistics & Data Science. Dr. Gilbert has received several awards, including a Sloan Research Fellowship (2006), an NSF CAREER award (2006), the National Academy of Sciences Award for Initiatives in Research (2008), the Association of Computing Machinery (ACM) Douglas Engelbart Best Paper award (2008), the EURASIP Signal Processing Best Paper award (2010), and the SIAM Ralph E. Kleinman Prize (2013).
Dr. Gilbert's research interests include analysis, probability, discrete mathematics, and algorithms. She is especially interested in randomized algorithms with applications to harmonic analysis, signal and image processing, and massive datasets.
Website: https://annacgilbert.github.io/cv/
]]>This observation gives rise to a simple question: how can this be? We answer this question by taking a careful look at the field of hyperbolic graph representation learning as it stands today, and find that a number of results do not diligently present baselines, make faulty modelling assumptions when constructing algorithms, and use misleading metrics to quantify geometry of graph datasets. We take a closer look at each of these three problems, elucidate the issues, perform an analysis of methods, and introduce a parametric family of benchmark datasets to ascertain the applicability of (hyperbolic) graph neural networks.
Unfortunately, these problems are not specific to hyperbolic graph neural nets but are indicative of more fundamental problems in graph machine learning more generally. We show, surprisingly, that node features are oftentimes more-than-sufficient for many common graph benchmarks, breaking this critical assumption. When comparing against a well-tuned feature-only MLP baseline on seven of the most commonly used graph learning datasets, one gains little benefit from using graph structure on five datasets. We posit that these datasets do not benefit considerably from graph learning because the features themselves already contain enough graph information to obviate or substantially reduce the need for the graph.
]]>From Democratizing Optimization with LLM to Improving LLM Performance with OR Techniques
Abstract:
Recent breakthroughs in Large Language Models (LLM) have captured public imagination and interest, while mathematical optimization remains largely underappreciated outside expert circles. In this talk, we show that LLM can finally bridge the persistent gap between optimization’s potent capabilities and its limited real-world uptake. We present the 4I framework—Insight, Interpretability, Interactivity, Improvisation—as a set of design principles for combining LLM with mathematical optimization. Insight establishes a trusted, up-to-date view of the state; Interpretability explains model logic and trade-offs; Interactivity enables conversational what-if analysis; and Improvisation supports event-driven re-optimization.
We also demonstrate how OR techniques can optimize LLM inference under memory constraints. We formulate LLM inference as a multi-stage online scheduling problem with stochastic arrivals and dynamic resource consumption. We develop a fluid dynamics approximation that provides a tractable performance benchmark and guides algorithm design. Building on this foundation, we introduce two algorithms. The Waiting for Accumulated Inference Threshold (WAIT) algorithm optimizes scheduling when output lengths are known using dynamically maintained thresholds. For the realistic scenario where output lengths are unknown at arrival, the Nested WAIT algorithm adaptively learns prompt characteristics through a hierarchical multi-segment framework. We establish theoretical near optimality guarantees under heavy traffic conditions, balancing throughput, latency, and Time to First Token (TTFT). Experiments using Llama-7B on A100 GPUs demonstrate 15-30% throughput improvements and reduced latency versus industry baselines (vLLM and Sarathi).
Bio:
David Simchi-Levi holds the MIT William Barton Rogers Professorship (named after the founder & first president of MIT), is a Professor of Engineering Systems at MIT and serves as the head of the MIT Data Science Lab. He is considered one of the premier thought leaders in supply chain management and business analytics. His Ph.D. students have accepted faculty positions in leading academic institutes including U. of California Berkeley, Carnegie Mellon U., Columbia U., Cornell U., Duke U., Georgia Tech, Harvard U., U. of Illinois Urbana-Champaign, U. of Michigan, Purdue U. and Virginia Tech.
Professor Simchi-Levi is the former Editor-in-Chief of Management Science (2018-2023), one of the two flagship journals of INFORMS. He served as the Editor-in-Chief for Operations Research (2006-2012), the other flagship journal of INFORMS and for Naval Research Logistics (2003-2005). In 2023, he was elected a member of the National Academy of Engineering. In 2020, he was awarded the prestigious INFORMS Impact Prize for playing a leading role in developing and disseminating a new highly impactful paradigm for the identification and mitigation of risks in global supply chains. He is an INFORMS Fellow and MSOM Distinguished Fellow and the recipient of the 2020 INFORMS Koopman Award given to an outstanding publication in military operations research; Ford Motor Company 2015 Engineering Excellence Award; 2014 INFORMS Daniel H. Wagner Prize for Excellence in Operations Research Practice; 2014 INFORMS Revenue Management and Pricing Section Practice Award; and 2009 INFORMS Revenue Management and Pricing Section Prize.
He was the founder of LogicTools which provided software solutions and professional services for supply chain optimization. LogicTools became part of IBM in 2009. In 2012 he co-founded OPS Rules, an operations analytics consulting company. The company became part of Accenture in 2016. In 2014, he co-founded Opalytics, a cloud analytics platform company focusing on operations and supply chain decisions. The company became part of the Accenture Applied Intelligence in 2018.
]]>We also demonstrate how OR techniques can optimize LLM inference under memory constraints. We formulate LLM inference as a multi-stage online scheduling problem with stochastic arrivals and dynamic resource consumption. We develop a fluid dynamics approximation that provides a tractable performance benchmark and guides algorithm design. Building on this foundation, we introduce two algorithms. The Waiting for Accumulated Inference Threshold (WAIT) algorithm optimizes scheduling when output lengths are known using dynamically maintained thresholds. For the realistic scenario where output lengths are unknown at arrival, the Nested WAIT algorithm adaptively learns prompt characteristics through a hierarchical multi-segment framework. We establish theoretical near optimality guarantees under heavy traffic conditions, balancing throughput, latency, and Time to First Token (TTFT). Experiments using Llama-7B on A100 GPUs demonstrate 15-30% throughput improvements and reduced latency versus industry baselines (vLLM and Sarathi).
]]>We strongly encourage you to attend to seek full-time employment, internships, and projects (rather than waiting until the end of the semester).
If you are interested in hosting a table for our upcoming session, please let us know after reviewing the below information within our website. Early registration closes December 1st!
Visit https://www.scl.gatech.edu/outreach/supplychainday.
]]>
Tirthankar Dasgupta is a Professor and Co-Director of Graduate Studies in the Department of Statistics, Rutgers University. His research interests include experimental design, causal inference, statistical modeling of physical and engineering systems and quality engineering. He obtained his Ph.D. in industrial engineering with specialization in engineering statistics from Georgia Institute of Technology in 2007 under the supervision of Prof C. F. Jeff Wu and received the Sigma-Xi best doctoral thesis award for his dissertation. He joined Harvard University in 2008 as an Assistant Professor and was promoted to the rank of Associate Professor in 2012. In the same year, he received the David Pickard award from Harvard University for excellence in teaching and mentoring. Prof Dasgupta has served as the research advisor of seven doctoral students till date. Part of his research has been funded by the Division of Civil, Mechanical and Manufacturing Innovation (CMMI), Division of Mathematical Sciences (DMS) and Division of the Social and Economic Sciences (SES) of the National Science Foundation. Currently he serves on the Editorial Boards of The Journal of the American Statistical Association and the Journal of the Royal Statistical Society (Series B).
This course focuses on defining, executing, and improving the sales and operations planning (S&OP) process. Participants will be introduced to the appropriate stakeholders of S&OP, the importance of S&OP to corporate performance, S&OP cadence, and the use of decision support tools to bring S&OP to the next level. Business cases will be used to show concrete examples of companies where S&OP is effectively applied.
This course is designed for chief operating officers; supply chain, sales, marketing and finance management executives (directors, vice presidents, executive vice presidents); supply chain and logistics managers, consultants, supervisors, planners, and engineers; supply chain education and human resource management personnel, inventory and demand planners, and procurement and sourcing analysts and managers; and manufacturing planners, analysts, and managers.
Gacha: A Simple Mechanism to Screen a Budget-Constrained Buyer
Abstract:
A lottery mechanism that allows repeated purchases until the buyer wins a designated item is widely used for both digital goods in Gacha games (e.g., Genshin Impact) and physical goods such as collectible toys (e.g., Labubu). We study this mechanism---referred to as a Gacha mechanism---in what is arguably the simplest possible setting: a seller offering one single item to a budget-constrained buyer. While the optimal mechanism in this setting typically involves designing a large (potentially infinite) menu of lotteries that can only be purchased once, the Gacha mechanism requires only a selling price and a winning probability, making it far more practical to implement. We show that the Gacha mechanism is particularly effective at screening along the budget dimension. While the posted price mechanism can perform arbitrarily poorly compared to the optimal mechanism when the buyer’s valuation is public, the Gacha mechanism achieves at least 63.2% of the optimal revenue. It becomes asymptotically optimal as the valuation grows large. Moreover, when the seller has almost no information about the buyer’s budget distribution, 63.2% is also the max-min revenue ratio guarantee, which can be achieved by a Gacha mechanism with vanishing winning probability. We also show that the Gacha mechanism is less effective at screening along the valuation dimension. When the valuation is private and the budget is public, the optimal Gacha mechanism reduces to a posted-price mechanism that achieves at least 50% of the optimal revenue. When both valuation and budget are private and independent, and the valuation follows a monotone hazard rate distribution, the Gacha mechanism guarantees at least 23% of the optimal revenue. Finally, we explore two extensions of the Gacha mechanism---one that includes the option of direct purchase, and another that incorporates a pity system.
Bio:
Zhenyu Hu is a Dean’s Chair Associate Professor at the Department of Analytics & Operations at the National University of Singapore. He obtained his PhD in industrial engineering from the University of Illinois at Urbana-Champaign and a bachelor in mathematics from Sun Yat-sen University. His research focuses on dynamic pricing and revenue management, supply chain management, and mechanism and information design. He currently serves as an associate editor for M&SOM.
]]>Thursday, December 4, 2025 | 12-1pm ET
We trust AI to make faster, smarter supply chain decisions. But can it be trusted to be ethical? Join Georgia Tech's Supply Chain and Logistics Institute and Rosemarie Santa Gonzalez to rethink what "ethical AI" really means in high-stakes environments like logistics, procurement, and risk management. This session will spark new thinking and help you move from AI ambition to AI accountability.
This course is the third in the four-course Supply Chain Analytics Professional certificate program. It introduces the field of machine learning, an area where algorithms learn patterns from data to support proactive decision making, as it applies to supply chain management. You’ll use machine learning to conduct predictive analytics as you forecast future demand, develop inventory policies, perform customer segmentation and predictive maintenance. You’ll use Python and PowerBI to create and analyze regression, clustering, and classification models.
The course is comprised of (4) half-day online instructor-led LIVE group webinars and pre-work (e.g. installing and testing software on your computer, testing connectivity with LMS and meeting software, etc.) to be completed before the first day of the course. An optional pre-course webinar is typically held the Thursday before the course start date (July 6).
Experienced business professionals who perform or want to perform analytics to improve their supply chain management processes. They want to tackle strategic goals and to perform leading edge analytics projects that address the full complexity of supply chains.
Abstract:
We introduce and analyze new probabilistic strategies for enforcing sufficient decrease conditions in stochastic derivative-free optimization, with the goal of reducing sample complexity and simplifying convergence analysis. First, we develop a new tail bound condition imposed on the estimated reduction in function value, which permits flexible selection of the power used in the sufficient decrease test, q in (1,2]. This approach allows us to reduce the number of samples per iteration from the standard O(delta^{−4}) to O(delta^{-2q}), assuming that the noise moment of order q/(q-1) is bounded. Second, we formulate the sufficient decrease condition as a sequential hypothesis testing problem, in which the algorithm adaptively collects samples until the evidence suffices to accept or reject a candidate step. This test provides statistical guarantees on decision errors and can further reduce the required sample size, particularly in the Gaussian noise setting, where it can approach O(delta^{−2-r}) when the decrease is of the order of delta^r. We incorporate both techniques into stochastic direct-search and trust-region methods for potentially non-smooth, noisy objective functions, and establish their global convergence rates and properties.
Bio:
Luis Nunes Vicente is the Timothy J. Wilmott ’80 Endowed Faculty Professor and Chair of Lehigh University’s Department of Industrial and Systems Engineering (ISE). His research interests include Continuous Optimization, Computational Science and Engineering, and Machine Learning and Data Science. He obtained his PhD from Rice University in 1996, under a Fulbright scholarship, receiving from Rice the Ralph Budd Thesis Award. He was one of the three finalists of the 94-96 A. W. Tucker Prize of the Mathematical Optimization Society (MOS). In 2015, he was awarded the Lagrange Prize of SIAM (Society for Industrial and Applied Mathematics) and MOS for the co-authorship of the book “Introduction to Derivative-Free Optimization, MPS-SIAM Series on Optimization, SIAM, Philadelphia, 2009”. He is a SIAM Fellow (Class of 2024). He was elected chair of the SIAM Activity Group on Optimization for 2023-2025 and President of the Association of Chairs of Operations Research Departments (ACORD) at INFORMS for 2024-2025. He has been chairing Lehigh ISE since August 2018.
]]>Often the lack of cooperation and coordination between organizations or stakeholders lead to inefficiencies, despite having common goals. A systems view is needed to ensure appropriate use of scarce resources to meet the multiple, and often conflicting, short- and long-term goals from multiple constituents. This course will focus on conceptual and modeling skills to understand and effectively manage supply chains and operations from a systems perspective. Models will address system characteristics (e.g., demand dependencies) that drive system dynamics and policies to regulate performance. Course topics include methods for improving coordination and collaboration, addressing demand dependencies, and reliably measuring and evaluating system performance.
This course is designed for representatives from governmental or non-governmental organizations, private corporations, military, and foundations, including but not limited to senior executives overseeing administrative and operational functions of an organization, logistics and supply chain managers, program managers, directors of field operations, directors of emergency/disaster preparedness and response, and public health professionals.
This course is the first in a 3-part virtually synchronous professional education program. Register and pay for all three required Health and Humanitarian Supply Chain Management Certificate courses and receive a discount of $400 off per course. Enter coupon code SCL-HHS at checkout with the Georgia Tech Professional Education website..
Additionally, there are scholarships available for the certificate program. Apply at https://chhs.gatech.edu/course-scholarships by the noted deadline.
Questions? Reach out to chhs@gatech.edu!
]]>Meeting demand in a timely and cost-effective manner is important both in public and private supply chains, and heavily depend on the design and operation of these supply chains. Demand is affected by ongoing factors such as local economy, infrastructure, and geographic location, as well as unexpected events such as natural or manmade disasters or other large-scale disruptions. Designing and operating responsive supply chains requires the consideration of uncertainty in timing, scope, scale, and understanding of various topics such as forecasting, distribution network design, and inventory management. This course will examine methods and models for making supply chain design and operational decisions and explore the significant value that is obtained through informed decision-making in advance of an unpredictable event or long-term strategy for meeting the need of customers and beneficiaries.
This course is designed for representatives from governmental or non-governmental organizations, private corporations, military, and foundations, including but not limited to senior executives overseeing administrative and operational functions of an organization, logistics and supply chain managers, program managers, directors of field operations, directors of emergency/disaster preparedness and response, and public health professionals.
This course is the first in a 3-part virtually synchronous professional education program. Register and pay for all three required Health and Humanitarian Supply Chain Management Certificate courses and receive a discount of $400 off per course. Enter coupon code SCL-HHS at checkout with the Georgia Tech Professional Education website..
Additionally, there are scholarships available for the certificate program. Apply at https://chhs.gatech.edu/course-scholarships by the noted deadline.
Questions? Reach out to chhs@gatech.edu!
]]>Generative AI Enabled Semantic Communication
Semantic communication (SemCom), a prominent feature of 6G, aims to address communication problems at the semantic level by transferring semantic information accurately and efficiently. Advances in generative artificial intelligence (GAI), such as the development of large language models and improved generative capabilities, have significantly facilitated the implementation of SemCom. This talk presents three cases of GAI empowering SemCom: The first case is a Swin-Transformer-based dynamic SemCom system that optimizes semantic communication efficiency by dynamically adjusting the compression rate based on network conditions for multi-user scenarios with varying computing capacities. The second case is a federated learning framework designed to enhance global model performance in decentralized environments by leveraging Federated Local Loss (FedLol) for efficient aggregation, reduced communication overhead, and effective image reconstruction. The third case is an AI-generated content framework (AIGC-SCM) for remote monitoring, utilizing GAI to achieve high-fidelity reconstruction of compressed content while maintaining semantic consistency and optimizing energy efficiency. Experimental results and demo confirm the effectiveness of these methods and provide practical insights for integrating SemCom with GAI.
Zhu Han received the B.S. degree in electronic engineering from Tsinghua University, in 1997, and the M.S. and Ph.D. degrees in electrical and computer engineering from the University of Maryland, College Park, in 1999 and 2003, respectively. From 2000 to 2002, he was an R&D Engineer of JDSU, Germantown, Maryland. From 2003 to 2006, he was a Research Associate at the University of Maryland. From 2006 to 2008, he was an assistant professor at Boise State University, Idaho. Currently, he is a John and Rebecca Moores Professor in the Electrical and Computer Engineering Department as well as the Computer Science Department at the University of Houston, Texas. Dr. Han is an NSF CAREER award recipient of 2010, and the winner of the 2021 IEEE Kiyo Tomiyasu Award. He has been an IEEE fellow since 2014, an AAAS fellow since 2020, ACM fellow since 2024, an IEEE Distinguished Lecturer from 2015 to 2018, and an ACM Distinguished Speaker from 2022-2025. Dr. Han is also a 1% highly cited researcher since 2017.
]]>Semantic communication (SemCom), a prominent feature of 6G, aims to address communication problems at the semantic level by transferring semantic information accurately and efficiently. Advances in generative artificial intelligence (GAI), such as the development of large language models and improved generative capabilities, have significantly facilitated the implementation of SemCom. This talk presents three cases of GAI empowering SemCom: The first case is a Swin-Transformer-based dynamic SemCom system that optimizes semantic communication efficiency by dynamically adjusting the compression rate based on network conditions for multi-user scenarios with varying computing capacities. The second case is a federated learning framework designed to enhance global model performance in decentralized environments by leveraging Federated Local Loss (FedLol) for efficient aggregation, reduced communication overhead, and effective image reconstruction. The third case is an AI-generated content framework (AIGC-SCM) for remote monitoring, utilizing GAI to achieve high-fidelity reconstruction of compressed content while maintaining semantic consistency and optimizing energy efficiency. Experimental results and demo confirm the effectiveness of these methods and provide practical insights for integrating SemCom with GAI.
]]>The requirement for high levels of customer service, increasing numbers of Stock Keeping Units (SKUs), and high labor costs have dramatically increased the complexity of warehouse operations. It is no longer sufficient to manage a warehouse based on a simple, arbitrary “ABC” classification of SKUs, which treats all those in a category as if they were identical. Instead, each decision, such as where to store or where to pick product, must be based on careful engineering and economic analysis.
This course is designed for supply chain and logistics consultants, supply chain engineers and analysts, facility engineers, and warehouse supervisors and team leaders.
Abstract:
Geometric qualification of a product is typically performed by specifying features or regions of interest (ROIs) during design, conducting shape registration to establish correspondence between the inspected product and its design counterpart, and measuring discrepancies for compliance assessment. For complex freeform products, the qualification often requires human intervention to ensure accuracy, particularly in personalized manufacturing through 3D printing. However, geometric variety and complexity can induce operator-to-operator variability due to heterogeneous spatial distributions of geometric distortions. To enable automated product qualification, we propose to specify ROIs as surface patches defined by geometric descriptors indicative of intrinsic deviation patterns. Utilizing these descriptors, ROI specification via shape space dimension reduction, non-rigid intrinsic shape registration, and intrinsic deviation representation can therefore be conducted automatically for product qualification. Finite types of ROIs or surface patches can be extracted based on their intrinsic deviation patterns, independent of covariates such as size and location. A software demo has been developed to implement the qualification process.
Bio:
Dr. Qiang Huang is a Professor at the Daniel J. Epstein Department of Industrial and Systems Engineering, University of Southern California (USC), Los Angeles. His research, detailed in his monograph "Domain-informed Machine Learning for Smart Manufacturing", has been focusing on machine learning for smart manufacturing and quality control for personalized manufacturing. He is an IISE Fellow Award, ASME Fellow, and a senior member of US National Academy of Inventors. He holds eight patents related to quality control in additive manufacturing.
]]>Abstract:
Network sequences are widely used to describe the longitudinal evolution of dynamic systems. Effective online monitoring of such sequences is crucial for detecting temporal structural changes in these systems. In the statistical process control (SPC) literature, a common approach is to extract key features from observed networks and then apply an SPC chart to monitor these features sequentially over time. However, existing methods often rely on features that are insensitive to certain important structural changes, and the control charts employed may not adequately capture the complex dependence structure among the extracted features. In our recent research, we propose four specific features to characterize the structure of an observed network. Collectively, these features can capture most of the structural changes of interest across various applications. After extracting these features, we employ a multivariate nonparametric control chart for online monitoring. Furthermore, we introduce a novel framework that represents each connected component of a network as a tree and the entire network as a forest. This forest-based representation enables intuitive 3D visualization and provides new structural features that enhance network comparison and monitoring. These new methods for network visualization and monitoring will be discussed in this talk, accompanied by extensive numerical demonstrations.
Bio:
Dr. Peihua Qiu is the Dean’s Professor and Founding Chair of the Department of Biostatistics at the University of Florida. He earned his PhD in Statistics from the University of Wisconsin–Madison in 1996. Dr. Qiu has made major contributions to several research areas, including jump regression analysis, image processing, statistical process control, survival analysis, dynamic disease screening, and spatio-temporal disease surveillance. He is the author of three books and more than 180 peer-reviewed journal articles. Dr. Qiu is an elected Fellow of the American Association for the Advancement of Science (AAAS), the American Statistical Association (ASA), the American Society for Quality (ASQ), and the Institute of Mathematical Statistics (IMS), and an elected member of the International Statistical Institute (ISI). He previously served as Editor of Technometrics (2014–2016) and is the 2024 recipient of the Shewhart Medal.
]]>Abstract:
Distributed and federated learning enables machine learning algorithms to be trained over decentralized edge devices without requiring the exchange of local datasets. We consider two scenarios in this talk. In the first scenario, we have mostly cooperative agents running distributed optimization methods. We analyze how the distribution of data affects agents' incentives to voluntarily participate and obediently follow traditional federated learning algorithms. We design a Faithful Federated Learning (FFL) mechanism based on FedAvg method and VCG mechanism which achieves (probably approximate) optimality, faithful implementation, voluntary participation, and balanced budget. We then analyze an alternative approach to align individual agent’s incentive to participate by allowing them to opt in or out. We propose a game theoretic framework and study the equilibrium properties with both rational and bounded rational agents. In the second scenario, we turn to a game theoretic formulation, where the agents may be under attack. We characterize the tradeoffs between convergence speed and robustness of learning dynamics.
Bio:
Ermin Wei is an Associate Professor at the Electrical and Computer Engineering Department and Industrial Engineering and Management Sciences Department of Northwestern University. She completed her PhD studies in Electrical Engineering and Computer Science at MIT in 2014, advised by Professor Asu Ozdaglar, where she also obtained her M.S. She received her undergraduate triple degree in Computer Engineering, Finance and Mathematics with a minor in German, from University of Maryland, College Park. Her team won the 2nd place in the GO-competition Challenge 1, an electricity grid optimization competition organized by Department of Energy. Wei's research interests include distributed optimization methods, convex optimization and analysis, smart grid, communication systems and energy networks and market economic analysis.
This course will demonstrate how warehouse operations are a key enabler to a successful supply chain implementation and the starting point for a transformation. It is critical to manage safety, quality and efficiency. Learn to leverage the lean supply chain modifications to improve customer responsiveness and reduce operating costs and in doing so contributing to a supply chain that creates a competitive advantage for a company. To accomplish this goal, we must bring lean principles into the warehouse and distribution center.
Who Should Attend
Supply chain professionals, logistics professionals, material managers, production control managers, transportation managers, warehousing managers and purchasing managers
Upon completion of this course, you will be able to:
Benefits:
Summary:
This seminar examines why exponential quantum advantages for NP-hard discrete optimization remain elusive and explores potential explanations and workarounds through the Maximum Cut problem.
Abstract:
Quantum computing offers hope for realizing exponential advantages over classical computing. Despite over two decades of work studying quantum approaches for discrete optimization, rigorously provable exponential advantages for approximating optimal solutions to NP-hard problems remain scarce. Why? We will discuss a potential explanation and workarounds, using the Maximum Cut problem as a running example.
Speaker Bio:
Ojas Parekh is a theoretical computer scientist who enjoys applying mathematical techniques to practically motivated interdisciplinary problems. He has worked in a variety of fields including discrete optimization, combinatorics, combinatorial scientific computing, and most recently, quantum and neuromorphic computing. A recent passion is helping shape the emerging fields of quantum approximation algorithms and quantum discrete optimization. He co-directs the Quantum Algorithms and Applications Collaboratory (QuAAC) at Sandia National Laboratories and directs the Department of Energy Fundamental Algorithmic Research toward Quantum Utility project (far-qu.sandia.gov), a multi-institutional effort tasked with designing novel quantum algorithms to realize advantages over classical computation, especially for optimization, simulation, and machine learning. Ojas believes the universe loves us because pizza exists.
Thursday, November 6, 2025 | 12-1pm ET
New Applications and Algorithms of Distributionally Robust Optimization in AI
In this talk, I will present our recent research on new applications and algorithms of Distributionally Robust Optimization (DRO) in AI, with a particular focus on training large foundation models such as contrastive language–image pretraining (CLIP) models. I will introduce a new learning framework, DRRHO risk minimization, which leverages open-weight models to accelerate the training of target models on custom datasets, and demonstrate its application to CLIP. By formulating the problem as a new class of finite-sum coupled compositional optimization, I will discuss how to design efficient algorithms with provable convergence guarantees. Finally, I will highlight the broader applications of these techniques across machine learning and AI.
In this talk, I will present our recent research on new applications and algorithms of Distributionally Robust Optimization (DRO) in AI, with a particular focus on training large foundation models such as contrastive language–image pretraining (CLIP) models. I will introduce a new learning framework, DRRHO risk minimization, which leverages open-weight models to accelerate the training of target models on custom datasets, and demonstrate its application to CLIP. By formulating the problem as a new class of finite-sum coupled compositional optimization, I will discuss how to design efficient algorithms with provable convergence guarantees. Finally, I will highlight the broader applications of these techniques across machine learning and AI.
]]>This course provides a practical, streamlined approach to building organizational resilience to disruptions. In our hyper-connected world, disruption is the new normal. Businesses face a relentless onslaught of risks – from supply chain breakdowns to cyber attacks and market shifts. Yet too many organizations rely on little more than hope when it comes to managing these threats. Course participants will discover how to leverage both classic risk management tools and cutting-edge technologies to proactively identify, assess, and mitigate a wide range of risks. Learn how to embed resilience planning into your regular business processes, so you are prepared for the unexpected.
This course is ideal for supply chain planning professionals, risk management and identification specialists, industry and government officials who are responsible for critical infrastructure, and business leaders in any sector concerned with potential operational disruptions.
Summary:
Sparsely sampled crossed random effects models arise in product review data, with effects for customers crossed with effects for products. The settings have no balance and the least squares algebra grows as N^(3/2) or worse. For generalized linear mixed models (GLMMs) there is the further difficulty of a very high dimensional integral. For instance we consider a likelihood with an integral over D~700,000 random effects, using only N~5,000,000 observations. The usual Laplace approximation method evaluates the D dimensional integral using just one integration point, and there is uncertainty about whether that is reliable. The MLE is infeasible in this problem and has only recently been shown to be consistent (Jiang, 2013). For a probit model, we develop a composite likelihood approach based on computing D one dimensional integrals. It is very scalable and we prove consistency which might not hold for the Laplace based method.
This is based on joint work with Ruggero Bellio and Swarnadip Ghosh and Cristiano Varin.
Bio:
Art Owen is the Max H. Stein Professor of Statistics at Stanford University. He is best known for inventing the empirical likelihood and for developing and studying randomized quasi-Monte Carlo methods. His research interests are centered on ways to measure uncertainty and ways to sample. He is a fellow of the American Statistical Association and the Institute of Mathematical Statistics. He received the 2020 Senior Noether Prize in nonparametric statistics from the ASA and the 2021 Gold Medal from the Statistical Society of Canada and became a SIAM Fellow in 2024.
]]>This is based on joint work with Ruggero Bellio and Swarnadip Ghoshand Cristiano Varin.
Many Supply Chain decisions are concerned with the timely and efficient procurement, allocation, and distribution of resources (e.g. funds, supplies, volunteers, money, employees) through a supply chain network. This course will explore methodologies for “medium term” decision making including procurement and inventory policies, strategies for distribution and allocation of limited resources, and supply chain design.
This course is designed for representatives from governmental or non-governmental organizations, private corporations, military, and foundations, including but not limited to senior executives overseeing administrative and operational functions of an organization, logistics and supply chain managers, program managers, directors of field operations, directors of emergency/disaster preparedness and response, and public health professionals.
This course is the second in a 3-part virtually synchronous professional education program. Register and pay for all three required Health and Humanitarian Supply Chain Management Certificate courses and receive a discount of $400 off per course. Enter coupon code SCL-HHS at checkout with the Georgia Tech Professional Education website..
Additionally, there are scholarships available for the certificate program. Apply at https://chhs.gatech.edu/course-scholarships by the noted deadline.
Questions? Reach out to chhs@gatech.edu!
]]>Often the lack of cooperation and coordination between organizations or stakeholders lead to inefficiencies, despite having common goals. A systems view is needed to ensure appropriate use of scarce resources to meet the multiple, and often conflicting, short- and long-term goals from multiple constituents. This course will focus on conceptual and modeling skills to understand and effectively manage supply chains and operations from a systems perspective. Models will address system characteristics (e.g., demand dependencies) that drive system dynamics and policies to regulate performance. Course topics include methods for improving coordination and collaboration, addressing demand dependencies, and reliably measuring and evaluating system performance.
This course is designed for representatives from governmental or non-governmental organizations, private corporations, military, and foundations, including but not limited to senior executives overseeing administrative and operational functions of an organization, logistics and supply chain managers, program managers, directors of field operations, directors of emergency/disaster preparedness and response, and public health professionals.
This course is the first in a 3-part virtually synchronous professional education program. Register and pay for all three required Health and Humanitarian Supply Chain Management Certificate courses and receive a discount of $400 off per course. Enter coupon code SCL-HHS at checkout with the Georgia Tech Professional Education website..
Additionally, there are scholarships available for the certificate program. Apply at https://chhs.gatech.edu/course-scholarships by the noted deadline.
Questions? Reach out to chhs@gatech.edu!
]]>Meeting demand in a timely and cost-effective manner is important both in public and private supply chains, and heavily depend on the design and operation of these supply chains. Demand is affected by ongoing factors such as local economy, infrastructure, and geographic location, as well as unexpected events such as natural or manmade disasters or other large-scale disruptions. Designing and operating responsive supply chains requires the consideration of uncertainty in timing, scope, scale, and understanding of various topics such as forecasting, distribution network design, and inventory management. This course will examine methods and models for making supply chain design and operational decisions and explore the significant value that is obtained through informed decision-making in advance of an unpredictable event or long-term strategy for meeting the need of customers and beneficiaries.
This course is designed for representatives from governmental or non-governmental organizations, private corporations, military, and foundations, including but not limited to senior executives overseeing administrative and operational functions of an organization, logistics and supply chain managers, program managers, directors of field operations, directors of emergency/disaster preparedness and response, and public health professionals.
This course is the first in a 3-part virtually synchronous professional education program. Register and pay for all three required Health and Humanitarian Supply Chain Management Certificate courses and receive a discount of $400 off per course. Enter coupon code SCL-HHS at checkout with the Georgia Tech Professional Education website..
Additionally, there are scholarships available for the certificate program. Apply at https://chhs.gatech.edu/course-scholarships by the noted deadline.
Questions? Reach out to chhs@gatech.edu!
]]>(*If availability doesn’t match, participants can select next best day for their schedule)
]]>]]>
]]>
ISyE Statistic Seminar – Xiaotong Shen
Synthetic data generation is reshaping data science by addressing challenges of scarcity, privacy, and imbalance. Recent advances in generative modeling enable the creation of high-fidelity datasets that capture complex distributions across modalities. These models not only expand data volume but also improve prediction accuracy, often outperforming conventional predictive methods, and can serve as a resampling tool for inference, much like the bootstrap. Moreover, they enhance multimodal analysis and provide targeted data augmentation for applied problems such as class imbalance.
Through case studies in sentiment analysis, predictive modeling, and tabular inference, we demonstrate how generative models enrich supervised learning, strengthen statistical inference, and provide scalable solutions when raw data are limited or biased.
This work is joint with Y. Liu, R. Shen, and X. Tian.
Xiaotong T. Shen is the John Black Johnston Distinguished Professor in the College of Liberal Arts at the University of Minnesota. He earned his Ph.D. in Statistics from the University of Chicago in 1991.
Professor Shen specializes in machine learning and data science, high-dimensional inference, non/semi-parametric inference, causal relations, graphical models, explainable Machine Intelligence (MI), personalization, recommender systems, natural language processing, generative modeling, and nonconvex minimization. His current research efforts are devoted to further developing causal and constrained inference, generative inference and prediction for black-box learners, and diffusion, normalizing flows, and summarization. The targeted application areas are biomedical sciences, artificial intelligence, and engineering.
He has served on the editorial boards of leading journals—including JMLR, JASA, and the Annals of Statistics—and remains deeply involved in professional service, currently as Chair-Elect of the ASA’s Section on Statistical Learning and Data Science. His distinctions include election as a Fellow of the Institute of Mathematical Statistics, the American Statistical Association, and AAAS, along with honors such as the “Scholar of the College” award and the ICSA Distinguished Achievement Award.
]]>Synthetic data generation is reshaping data science by addressing challenges of scarcity, privacy, and imbalance. Recent advances in generative modeling enable the creation of high-fidelity datasets that capture complex distributions across modalities. These models not only expand data volume but also improve prediction accuracy, often outperforming conventional predictive methods, and can serve as a resampling tool for inference, much like the bootstrap. Moreover, they enhance multimodal analysis and provide targeted data augmentation for applied problems such as class imbalance.
Through case studies in sentiment analysis, predictive modeling, and tabular inference, we demonstrate how generative models enrich supervised learning, strengthen statistical inference, and provide scalable solutions when raw data are limited or biased.
This work is joint with Y. Liu, R. Shen, and X. Tian.
]]>
Georgia Tech Exhibition Hall (2nd floor)
Centennial Room 9:15-10am Breakfast/Networking
Home Park Room 10-11am Main Seminar
Georgia Institute of Technology
460 4th St NW, Atlanta, GA 30332
Neil Cawse will share his insights on:
Neil Cawse, CEO and founder of Geotab, stands at the forefront of connected vehicle technology, leading the industry in innovation and research. His leadership is marked by a commitment to developing cutting-edge technologies that empower businesses to make data-driven decisions. Under Neil’s guidance, Geotab has transformed into a global leader, providing advanced data analytics and insights to businesses worldwide. Neil’s entrepreneurial spirit and passion for technology have been the driving forces behind Geotab's mission to improve vehicle safety, operational efficiency, and sustainability. With a keen focus on AI, IoT and connected vehicles, he has steered the company through exponential growth, expanding its reach to millions of subscriptions in over 160 countries. Neil's vision is to create a more connected and sustainable world through the power of data and innovation.
Additional speakers include: Mike Branch (Geotab VP of Data & Analytics), John Ballantyne (Geotab VP of Research and Innovation), Samuel Harris (GDOT Asst State Traffic Engineer), Benoit Montreuil (SCL Executive Director), Yichang (James) Tsai (Prof, Civil and Environmental Engr), Sofia Perez-Guzman (Asst Prof, Civil and Environmental Engr), Chris Gaffney (SCL Managing Director)
The panelists will review how Traditional AI has improved fleet operations in safety, maintenance, and efficiency. They will then examine how Generative and Agentic AI can automate aspects of fleet management. Finally, the discussion will shift to how transportation planners, operators, and policymakers can benefit from anonymized geotemporal data.
For questions, please email event@scl.gatech.edu.
Georgia Tech Exhibition Hall (2nd floor)
Centennial Room 9:15-10am Breakfast/Networking
Home Park Room 10-11am Main Seminar
Georgia Institute of Technology
460 4th St NW, Atlanta, GA 30332
Summary:
Classical statistical learning theory primarily concerns independent data. In comparison, it has been much less investigated for time dependent data, which are commonly encountered in economics, engineering, finance, geography, physics and other fields. In this talk, we focus on concentration inequalities for suprema of empirical processes which plays a fundamental role in the statistical learning theory. We derive a Gaussian approximation and an upper bound for the tail probability of the suprema under conditions on the size of the function class, the sample size, temporal dependence and the moment conditions of the underlying time series. Due to the dependence and heavy-tailness, our tail probability bound is substantially different from those classical exponential bounds obtained under the independence assumption in that it involves an extra polynomial decaying term. We allow both short- and long-range dependent processes, where the long-range dependence case has never been previously explored. We showed our tail probability inequality is sharp up to a multiplicative constant. These bounds work as theoretical guarantees for statistical learning applications under dependence.
Bio:
Wei Biao Wu earned his Ph.D. in Statistics from the University of Michigan, Ann Arbor, in 2001. He is a Professor in the Department of Statistics at The University of Chicago, Chicago, IL. His research focuses on developing large-sample theory for time-dependent data, introducing the innovative functional dependence measure to establish a robust theoretical framework for asymptotic analysis. He has also designed efficient online algorithms for computing time-series statistics. His research interests span probability theory, statistics, and econometrics.
]]>AI, BI & SI—Artificial, Biological and Statistical Intelligences
Artificial Intelligence (AI) is clearly one of the hottest subjects these days. Basically, AI employs a huge number of inputs (training data), super-efficient computer power/memory, and smart algorithms to perform its intelligence. In contrast, Biological Intelligence (BI) is a natural intelligence that requires very little or even no input. This talk will first discuss the fundamental issue of input (training data) for AI. After all, not-so-informative inputs (even if they are huge) will result in a not-so-intelligent AI. Specifically, three issues will be discussed: (1) input bias, (2) data right vs. right data, and (3) sample vs. population. Finally, the importance of Statistical Intelligence (SI) will be introduced. SI is somehow in between AI and BI. It employs important sample data, solid theoretically proven statistical inference/models, and natural intelligence. In my view, AI will become more and more powerful in many senses, but it will never replace BI. After all, it is said that “The truth is stranger than fiction, because fiction must make sense.” The ultimate goal of this study is to find out “how can humans use AI, BI, and SI together to do things better.”
Dr. Dennis K. J. Lin is a Distinguished Professor of Statistics at Purdue University. He served as the Department Head during 2020-2022. Prior to this current job, he was a University Distinguished Professor of Supply Chain Management and Statistics at Penn State, where he worked for 25 years. His research interests are data quality, industrial statistics, statistical inference, and data science. He has published nearly 300 SCI/SSCI papers in a wide variety of journals. He currently serves or has served as an associate editor for more than 10 professional journals and was a co-editor for Applied Stochastic Models for Business and Industry. Dr. Lin is an elected fellow of ASA, IMS, ASQ, & RSS, an elected member of ISI, and a lifetime member of ICSA. He is an honorary chair professor for various universities, including Fudan University, and National Taiwan Normal University and a Chang-Jiang Scholar at Renmin University of China,. His recent awards include, the Youden Address (ASQ, 2010), the Shewell Award (ASQ, 2010), the Don Owen Award (ASA, 2011), the Loutit Address (SSC, 2011), the Hunter Award (ASQ, 2014), the Shewhart Medal (ASQ, 2015), and the SPES Award (ASA, 2016). He was the Deming Lecturer Award at 2020 JSM. His most recent award is “The 2022 Distinguished Alumni Award” (National Tsing Hua University, Taiwan).
]]>Artificial Intelligence (AI) is clearly one of the hottest subjects these days. Basically, AI employs a huge number of inputs (training data), super-efficient computer power/memory, and smart algorithms to perform its intelligence. In contrast, Biological Intelligence (BI) is a natural intelligence that requires very little or even no input. This talk will first discuss the fundamental issue of input (training data) for AI. After all, not-so-informative inputs (even if they are huge) will result in a not-so-intelligent AI. Specifically, three issues will be discussed: (1) input bias, (2) data right vs. right data, and (3) sample vs. population. Finally, the importance of Statistical Intelligence (SI) will be introduced. SI is somehow in between AI and BI. It employs important sample data, solid theoretically proven statistical inference/models, and natural intelligence. In my view, AI will become more and more powerful in many senses, but it will never replace BI. After all, it is said that “The truth is stranger than fiction, because fiction must make sense.” The ultimate goal of this study is to find out “how can humans use AI, BI, and SI together to do things better.”
]]>This course is the third in the four-course Supply Chain Analytics Professional certificate program. It introduces the field of machine learning, an area where algorithms learn patterns from data to support proactive decision making, as it applies to supply chain management. You’ll use machine learning to conduct predictive analytics as you forecast future demand, develop inventory policies, perform customer segmentation and predictive maintenance. You’ll use Python and PowerBI to create and analyze regression, clustering, and classification models.
The course is comprised of (4) half-day online instructor-led LIVE group webinars and pre-work (e.g. installing and testing software on your computer, testing connectivity with LMS and meeting software, etc.) to be completed before the first day of the course. An optional pre-course webinar is typically held the Thursday before the course start date (July 6).
Experienced business professionals who perform or want to perform analytics to improve their supply chain management processes. They want to tackle strategic goals and to perform leading edge analytics projects that address the full complexity of supply chains.
Every fall and spring semester, during the IISE Career Fair, companies across the nation come to Georgia Tech to recruit some of the nation’s top talent from our Bachelor’s and Master’s programs. Our students are recruited for a variety of roles and perform well past expectations in all positions. We hope that you will join us this semester and meet some of the country’s brightest students.
This fall semester, we will be hosting an in-person career fair on Tuesday, September 2nd, 2025, at Georgia Tech’s Exhibition Hall. We highly encourage all companies to register early for the Fall 2025 Career Fair. The first 5 companies to register and pay will have the exclusive opportunity to reserve a dedicated room for additional networking after the event. This private networking session can be tailored to your preferences, giving your company a unique chance to connect with top talent in a more intimate setting. Don't miss out on this valuable opportunity to enhance your recruitment experience!
Join us at one of the largest Industrial & Systems Engineering career fairs in the nation. Registration closes August 27th, so secure your table today!:
Looking for internships or full-time roles? This is your chance to meet leading employers and explore exciting career opportunities in Industrial & Systems Engineering. For more information and to submit your resume, visit https://www.gtiise.org/career-fair-info.
]]>Abstract:
There has been significant developments of large and complex engineered infrastructure systems such as telecommunication networks, power grids, transportation infrastructure, healthcare delivery systems, information technology, financial systems and supply chain systems. Failures of such systems may result in cascading damages as well as significant interruptions of their services. This presentation focuses on the development of unified resilience metrics based on analytical and AI-enabled stochastic models to evaluate and predict the resilience and availability of complex systems subject to various types of failures, repairs and recovery processes.
Bio:
E. A. Elsayed is Distinguished Professor of the Department of Industrial and Systems Engineering, Rutgers University. He is also the Director of the NSF/ Industry/ University Co-operative Research Center for Quality and Reliability Engineering. He was the Chair of ISE, Rutgers University from 1983 to 2001. His research interests are in the areas of quality and reliability engineering and Production Planning and Control. He is a co-author of Quality Engineering in Production Systems, McGraw Hill Book Company, 1989. He is also the author of Reliability Engineering, Addison-Wesley, 1996. These two books received the 1990 and 1997 IIE Joint Publishers Book-of-the-Year Award respectively. His recent book Reliability Engineering 2nd Edition, Wiley, 2021 received the 2012 Outstanding Publications of IIE. The third edition of this book by Wiley in 2021 was selected to be included in the Best Industrial Management eBooks of All Time. Dr. Elsayed has received many awards and honors and was the keynote speaker of many international conferences. Dr. Elsayed was awarded the Doctor Honoris Causes from University of Agers, France in January 2018 for his achievements in the reliability engineering field. In November 2023, he received “Faculty of the Year” award, Rutgers University.
]]>Thursday, August 7, 2025 | 12-1pm ET
An educational webinar where we explore strategies for unlocking business potential through integrated alignment of operations, supply chain, and finance. In this expert-led session, faculty from the Scheller College of Business and Stewart School of Industrial and Systems Engineering guide you through the fundamentals of Integrated Business Planning (IBP), demonstrate its power to transform enterprise performance, and share practical insights for leading organizational change.
Thursday, October 2, 2025 | 12-1pm ET
In today’s volatile business landscape, resilience isn’t just a buzzword, it’s a competitive necessity. Join the Georgia Tech Supply Chain and Logistics Institute to explore how project management strategies can turn unpredictable supply chain environments into opportunities for stability and growth. Whether you're leading cross-functional initiatives, managing vendor transitions, or navigating disruption, you’ll learn why traditional approaches fall short and how integrated project and supply chain thinking can help you deliver lasting results. Perfect for project managers, supply chain professionals, and anyone tasked with delivering complex initiatives in dynamic settings.
Abstract:
Sampling from densities defined on Riemannian manifolds is central to Bayesian inference, generative modeling, and differential privacy. We introduce the Riemannian Proximal Sampler (RPS), whose efficiency hinges on two oracles: Manifold Brownian Increments and the Riemannian Heat Kernel. We establish high-accuracy sampling guarantees for the Riemannian Proximal Sampler, showing that generating samples with ε-accuracy requires O(log(1/ε)) iterations in Kullback-Leibler divergence assuming access to exact oracles and O(log^2(1/ε)) iterations in the total variation metric assuming access to sufficiently accurate inexact oracles. Furthermore, we present two practical implementations of these oracles by leveraging heat-kernel truncation and Varadhan's asymptotics, respectively. In the latter case, we interpret the Riemannian Proximal Sampler as a discretization of the entropy-regularized Riemannian Proximal Point Method on the associated Wasserstein space. We will discuss numerical results that illustrate the effectiveness of the proposed methodology.
Bio:
Krishnakumar Balasubramanian is an Associate Professor in the Department of Statistics at the University of California, Davis, affiliated with the Graduate Group in Applied Mathematics, the Center for Data Science and Artificial Intelligence Research (CeDAR), and the TETRAPODS Institute of Data Science. He is also an Amazon Scholar and was a visiting scientist at the Simons Institute for the Theory of Computing at UC Berkeley in Fall 2021 and Fall 2022. Krishna received his Ph.D. in Computer Science from the Georgia Institute of Technology and completed postdoctoral research at Princeton University and the University of Wisconsin–Madison. His research lies at the interface of machine learning and artificial intelligence, statistics and optimization. He is a recipient of several honors, including a Facebook Fellowship (2013), the ICML Best Paper Runner-Up Award (2013), and the INFORMS ICS Prize (2024). He contributes actively to the academic community as an Associate Editor for the Annals of Statistics, IEEE Transactions on Information Theory and the Journal of Machine Learning Research, and serves regularly as a (senior) area chair for leading conferences such as ICML, ICLR, NeurIPS, and COLT.
]]>Abstract:
Technological advances in digital platforms, automation, and sensor systems are rapidly expanding capabilities in healthcare delivery and humanitarian assistance. However, in complex operating environments, the integration of these technologies often reveals a critical gap between cyber systems and the physical realities of human action, environmental variability, and contextual constraints. This misalignment can undermine decision quality and service reliability, and in some cases, lead to unintended or suboptimal outcomes.
This talk examines these challenges in last-mile delivery for healthcare and humanitarian assistance, where decision-making is shaped by the dynamic interplay of human judgment, technological capabilities, and on-the-ground realities. Drawing on case studies, it will discuss how discrepancies between data collected for one operational purpose and its subsequent use for modeling or strategic decision-making can propagate through the system—amplifying risks and degrading performance. Emphasis will be placed on the importance of context-aware design, stakeholder engagement, and research translation to deliver high-quality, sustainable solutions.
Bio:
Dr. Yuehwern Yih is the Tompkins Professor of Industrial Engineering at Purdue University. She previously served as the Director of LASER PULSE ($70 million USAID funded program) and the Associate Director of Regenstrief Center for Healthcare Engineering. She is a Member of the National Academy of Engineering (NAE) a Senior Member of the National Academy of Inventors (NAI), an Institute of Industrial and Systems Engineers (IISE) Fellow, and an Executive Leadership in Academic Technology, Engineering and Science (ELATE) Fellow.
Dr. Yih’s core research focuses on understanding dynamics of system behaviors to improve the performance of complex systems under volatile environments including manufacturing systems, supply chains, humanitarian assistance, health care delivery, and global development. Dr. Yih received the IISE David F Baker Distinguished Research Award, the Melinda and Bill Gates Grand Challenge Award, the inaugural Faculty Engagement Fellow (highest honor for engagement at Purdue), and multiple Pritsker Undergraduate Teaching Awards and the Most Impactful Faculty Inventors at Purdue. She has vast experience in interdisciplinary and cross-sector collaboration to address global development challenges, e.g. integrated nutrition system for HIV patients in Kenya, medical supply chains for maternal health in Uganda, humanitarian supply chains in South Sudan and Ukraine, impacting over a million people in need. Dr. Yih received her B.S. in Industrial Engineering from the National Tsing Hua University in Taiwan, and her Ph.D. from the University of Wisconsin-Madison.
]]>This talk examines these challenges in last-mile delivery for healthcare and humanitarian assistance, where decision-making is shaped by the dynamic interplay of human judgment, technological capabilities, and on-the-ground realities. Drawing on case studies, it will discuss how discrepancies between data collected for one operational purpose and its subsequent use for modeling or strategic decision-making can propagate through the system—amplifying risks and degrading performance. Emphasis will be placed on the importance of context-aware design, stakeholder engagement, and research translation to deliver high-quality, sustainable solutions.
]]>The requirement for high levels of customer service, increasing numbers of Stock Keeping Units (SKUs), and high labor costs have dramatically increased the complexity of warehouse operations. It is no longer sufficient to manage a warehouse based on a simple, arbitrary “ABC” classification of SKUs, which treats all those in a category as if they were identical. Instead, each decision, such as where to store or where to pick product, must be based on careful engineering and economic analysis.
This course is designed for supply chain and logistics consultants, supply chain engineers and analysts, facility engineers, and warehouse supervisors and team leaders.
Thursday, September 4, 2025 | 12-1pm ET
Expect a fast-paced, interactive format featuring real demos, “under-the-hood” views of emerging trends like agents, and our new Advanced Analytics Learning Ladder—an actionable guide to upskilling individuals and teams from fundamentals to frontier. Whether you’re attending in person or online, you’ll leave with tools and insights you can apply immediately to reclaim time, boost productivity, and future-proof your supply chain talent strategy.
Featured Speaker: Chris Gaffney, Managing Director of the Supply Chain and Logistics Institute and Edenfield Executive-in-Residence, and a Professor of the Practice.
The event will be hosted at Georgia Tech Savannah, but note that this is a hybrid event (attendees can join in-person or virtually, but you must register to attend).
Cost: Free
We strongly encourage you to attend to seek full-time employment, internships, and projects (rather than waiting until the end of the semester).
If you are interested in hosting a table for the upcoming session, please let us know after reviewing the below information within our website. Early registration closes August 1st!
Visit https://www.scl.gatech.edu/outreach/supplychainday.
]]>The requirement for high levels of customer service, increasing numbers of SKUs and high labor costs have dramatically increased the complexity of warehouse operations. It is no longer sufficient to manage a warehouse based on a simple, arbitrary “ABC” classification of SKUs, which treats all those in a category as if they were identical. Instead, each decision – such as where to store or where to pick product – must be based on careful engineering and economic analysis. Each SKU must identify its own cheapest, fastest path through the warehouse to the customer and then compete with all the other SKUs for the necessary resources. This results in warehouse operations that are finely tuned to patterns of customer orders and maximally efficient. Learn the concepts necessary to address modern warehouse trade-offs between space and time in optimizing and managing your warehouse.
Essential learning for those who are seeking cost reductions through better handling methods. Also valuable for those who must replace, upgrade, or add material handling equipment. The two-day course will include case examples and a guided exercise to ensure mastery of the techniques presented.
Supply chain and logistics consultants, supply chain engineers and analysts, facility engineers, and warehouse supervisors and team leaders
Upon completion of this course, you will be able to:
None.
This course is part of the Distribution Operations Analysis & Design (DOAD) Certificate.
]]>Thursday, July 3, 2025 | 12-1pm ET
This course is the fourth in the 4-course Supply Chain Analytics Professional certificate program. It incorporates learning advanced analytics and mathematical optimization to find solutions for supply chain problems. You’ll learn how to use linear programming, mixed integer programming, and heuristics to conduct prescriptive analytics related to production processes, distribution networks, and routing. The course serves as a capstone for the program by culminating in a hackathon where you’ll design networks, inventory policies, and scenarios and then evaluate the outcomes via simulations.
The online version of the course is comprised of (4) half-day online instructor-led LIVE group webinars and pre-work (e.g. installing and testing software on your computer, testing connectivity with LMS and meeting software, etc.) to be completed before the first day of the course.
Experienced business professionals who perform or want to perform analytics to improve their supply chain management processes. They want to tackle strategic goals and to perform leading edge analytics projects that address the full complexity of supply chains.
This course is the second in the four-course Supply Chain Analytics Professional certificate program. It emphasizes operational performance metrics to align supply chain management with strategic business goals. You’ll learn several statistics concepts (e.g. variance analysis, hypothesis testing, forecasting methods) along with inventory management models. You’ll use diagnostic analytics with PowerBI and Python to conduct demand and service profiling, undertake root cause analysis, and use time series forecasting in inventory management.
The online version of the course is comprised of (4) half-day online instructor-led LIVE group webinars and pre-work (e.g. installing and testing software on your computer, testing connectivity with LMS and meeting software, etc.) to be completed before the first day of the course.
Experienced business professionals who perform or want to perform analytics to improve their supply chain management processes. They want to tackle strategic goals and to perform leading edge analytics projects that address the full complexity of supply chains.
This course is the first in the four-course Supply Chain Analytics Professional certificate program. It prepares you to apply leading-edge analytical methods and technology enablers across the supply chain. You’ll learn the dynamics of supply chains, the most relevant planning challenges, and the roles of different types of analytics. Next, you’ll learn about data cleansing, exploratory data analysis, and visualization. You’ll use Python and PowerBI to analyze the causes of underperformance and to build dashboards to visualize supply chain data. You will leave knowing how to gather, analyze, and prepare your data through descriptive analytics before you dig into deeper applications.
The online version of the course is comprised of (4) half-day instructor-led LIVE group webinars and pre-work (e.g. installing and testing software on your computer, testing connectivity with LMS and meeting software, etc.) to be completed before the first day of the course.
Experienced business professionals who perform or want to perform analytics to improve their supply chain management processes. They want to tackle strategic goals and to perform leading edge analytics projects that address the full complexity of supply chains.
Essentials of Negotiations and Stakeholder Influence level-sets the participants' understanding of negotiation influence and strengthens preparation, planning and execution activities involved with both simple and complex negotiations. The program includes industry techniques and tools for traditional supplier negotiations, as well as tips for internal cross-functional leadership. Participants walk away with a standard industry and customized individual experience which includes their personal Negotiation Style “DNA” to help them embrace their own natural tendencies and strengths. The program includes mock negotiations to reinforce techniques and tactics immediately in a “no judgement zone” environment.
This course is ideal for sourcing initiative leaders, project leaders, business unit leaders, operations managers, sales leaders and procurement & supply management-related professionals who are involved with supplier selection, contract development and supplier performance management.
Contracting and Legal Oversight provides participants with a holistic and integrated understanding of contract law, contract types, key industry standard contract terms, and contract structure to improve their confidence when creating or modifying contract documents. The program is geared to reinforce standards of excellence for professionals who are responsible for delivering contractual agreements and mitigating financial risk for their organization.
The online version of the course is comprised of (3) instructor-led LIVE group webinars, homework, and pre-work (e.g. installing and testing software on your computer, testing connectivity with Canvas LMS and BlueJeans meeting software, etc.) to be completed before the first day of the course.
This course is ideal for contract managers, procurement professionals, sourcing initiative leaders, project managers and all procurement & supply management-related professionals involved with bid contract development, contract execution or supplier performance management.
Category Management and Sourcing Leadership is designed to deepen participants' knowledge base of core activities in the procurement & supply management function. The program covers the sourcing process, specifications gathering, common bid package alternatives, cross-functional collaboration and supplier evaluation & selection. Participants will walk away ready to develop bid packages more thoroughly to help drive sourcing decisions for their organizations. This "hands on" delivery focuses on the professional serving as the main liaison between the buying organization and the selling organization in the company sourcing process.
This course is ideal for sourcing initiative leaders, procurement professionals, project managers, finance analyst, contract managers and all procurement & supply management-related professionals involved with bid package development, bid package analysis, negotiations preparation, contracting and supplier selection activity.
This course will demonstrate how warehouse operations are a key enabler to a successful supply chain implementation and the starting point for a transformation. It is critical to manage safety, quality and efficiency. Learn to leverage the lean supply chain modifications to improve customer responsiveness and reduce operating costs and in doing so contributing to a supply chain that creates a competitive advantage for a company. To accomplish this goal, we must bring lean principles into the warehouse and distribution center.
Who Should Attend
Supply chain professionals, logistics professionals, material managers, production control managers, transportation managers, warehousing managers and purchasing managers
Upon completion of this course, you will be able to:
Benefits:
This course focuses on defining, executing, and improving the sales and operations planning (S&OP) process. Participants will be introduced to the appropriate stakeholders of S&OP, the importance of S&OP to corporate performance, S&OP cadence, and the use of decision support tools to bring S&OP to the next level. Business cases will be used to show concrete examples of companies where S&OP is effectively applied.
This course is designed for chief operating officers; supply chain, sales, marketing and finance management executives (directors, vice presidents, executive vice presidents); supply chain and logistics managers, consultants, supervisors, planners, and engineers; supply chain education and human resource management personnel, inventory and demand planners, and procurement and sourcing analysts and managers; and manufacturing planners, analysts, and managers.
This course prepares you for the basics of transportation operations and analysis. Throughout the course, you’ll review the key elements of transportation such as modes of transportation, transportation procurement, cost minimization techniques, the role of ports in global logistics, and international trade terms. There will also be discussion around emerging trends in North American transportation markets, emerging techniques, and greenhouse gas emissions reduction.
This course is designed for Supply Chain Managers, Distribution Managers, Transportation Planners, Transportation Clerks, Transportation Analysts, and Transportation Managers and learners seeking to enter these roles. Supply chain professionals from other domains will also benefit through gaining insights into transportation operations.
This course prepares you for the basics of transportation operations and analysis. Throughout the course, you’ll review the key elements of transportation such as modes of transportation, transportation procurement, cost minimization techniques, the role of ports in global logistics, and international trade terms. There will also be discussion around emerging trends in North American transportation markets, emerging techniques, and greenhouse gas emissions reduction.
This course is designed for Supply Chain Managers, Distribution Managers, Transportation Planners, Transportation Clerks, Transportation Analysts, and Transportation Managers and learners seeking to enter these roles. Supply chain professionals from other domains will also benefit through gaining insights into transportation operations.
Thursday, May 1, 2025 | 12-1pm ET
Thursday, June 5, 2025 | 12-1pm ET
This course focuses on defining, executing, and improving the sales and operations planning (S&OP) process. Participants will be introduced to the appropriate stakeholders of S&OP, the importance of S&OP to corporate performance, S&OP cadence, and the use of decision support tools to bring S&OP to the next level. Business cases will be used to show concrete examples of companies where S&OP is effectively applied.
This course is designed for chief operating officers; supply chain, sales, marketing and finance management executives (directors, vice presidents, executive vice presidents); supply chain and logistics managers, consultants, supervisors, planners, and engineers; supply chain education and human resource management personnel, inventory and demand planners, and procurement and sourcing analysts and managers; and manufacturing planners, analysts, and managers.
This course provides a practical, streamlined approach to building organizational resilience to disruptions. In our hyper-connected world, disruption is the new normal. Businesses face a relentless onslaught of risks – from supply chain breakdowns to cyber attacks and market shifts. Yet too many organizations rely on little more than hope when it comes to managing these threats. Course participants will discover how to leverage both classic risk management tools and cutting-edge technologies to proactively identify, assess, and mitigate a wide range of risks. Learn how to embed resilience planning into your regular business processes, so you are prepared for the unexpected.
This course is ideal for supply chain planning professionals, risk management and identification specialists, industry and government officials who are responsible for critical infrastructure, and business leaders in any sector concerned with potential operational disruptions.
This course is the third in the four-course Supply Chain Analytics Professional certificate program. It introduces the field of machine learning, an area where algorithms learn patterns from data to support proactive decision making, as it applies to supply chain management. You’ll use machine learning to conduct predictive analytics as you forecast future demand, develop inventory policies, perform customer segmentation and predictive maintenance. You’ll use Python and PowerBI to create and analyze regression, clustering, and classification models.
The course is comprised of (4) half-day online instructor-led LIVE group webinars and pre-work (e.g. installing and testing software on your computer, testing connectivity with LMS and meeting software, etc.) to be completed before the first day of the course. An optional pre-course webinar is typically held the Thursday before the course start date (July 6).
Experienced business professionals who perform or want to perform analytics to improve their supply chain management processes. They want to tackle strategic goals and to perform leading edge analytics projects that address the full complexity of supply chains.
When Does Interference Matter? Decision-Making in Platform Experiments
Abstract:
Online platforms and marketplaces use A/B experiments to test new features and design changes. Due to constraints on inventory, such experiments typically lead to biased estimation of treatment effects due to the presence of *interference* between treatment and control groups; this phenomenon has been extensively studied in recent literature. By contrast, there has been relatively little discussion of the impact of interference on *decision-making*. In this talk, we consider a benchmark Markovian model of a capacity-constrained platform, and study the impact of interference on (1) false positive probability and (2) statistical power. We show that for a particular class of "monotone" treatments (informally, treatments where the sign of the effect does not depend on the level of available inventory), using the standard t statistic with the naïve difference-in-means estimator and classical variance estimator both correctly controls the false positive probability, and generally yields *higher* statistical power than any unbiased estimation method. We show that in principle, these effects can be undermined when treatments are not monotone.
Our results have important implications for the practical deployment of debiasing strategies for A/B experiments. In particular, they highlight the need for platforms to carefully define their objectives and understand the nature of their interventions when determining appropriate estimation and decision-making approaches. Notably, when interventions are monotone, the platform may actually be worse off by pursuing a debiased decision-making approach.
Joint work with Hannah Li, Anushka Murthy, and Gabriel Weintraub.
Bio:
Ramesh Johari is a Professor at Stanford University, with a full-time appointment in the Department of Management Science and Engineering (MS&E), and a courtesy appointment in the Department of Electrical Engineering (EE). He is an associate director of Stanford Data Science, and co-director of the Stanford Causal Science Center. He is a member of the Operations Research group and the Social Algorithms Lab (SOAL) in MS&E, the Information Systems Laboratory in EE, and the Institute for Computational and Mathematical Engineering. He received an A.B. in Mathematics from Harvard, a Certificate of Advanced Study in Mathematics from Cambridge, and a Ph.D. in Electrical Engineering and Computer Science from MIT. His current research interests include market design, causal inference, and experimentation.
]]>Our results have important implications for the practical deployment of debiasing strategies for A/B experiments. In particular, they highlight the need for platforms to carefully define their objectives and understand the nature of their interventions when determining appropriate estimation and decision-making approaches. Notably, when interventions are monotone, the platform may actually be worse off by pursuing a debiased decision-making approach.
Joint work with Hannah Li, Anushka Murthy, and Gabriel Weintraub.
]]>The Physical Internet Initiative aims at transforming the way physical objects are moved, stored, realized, supplied and used, pursuing global logistics efficiency and sustainability. Originating from Professor Benoit Montreuil in 2006, this ground breaking vision, revolutionizing current paradigms, has stirred great interest from scientific, industrial as well as governmental communities.
The International Physical Internet Conference aims to provide an open forum for researchers, industry representatives, government officials and citizens to together explore, discuss, introduce leading edge concepts, methodologies, recent projects, technological advancements, start-up initiatives, for current and future Physical Internet implementation.
Conference topics include inter-connected logistics, PI fundamentals, business models, governance and implementation, cross-chain control, synchromodal transportation, IT systems, stakeholders and their roles. New business models, enabling technologies and experimentations already underway will be presented, making this meeting a unique opportunity to learn, network and discuss the latest results and challenges about interconnected logistics.
And, because logistics is global, participants will be from all over the world including researchers, industrial and international institution members, local authorities and standardization committees.
]]>Please join us for the 11th International Physical Internet Conference taking place June 18-20, 2025 in Hong Kong, China.
This course will demonstrate how warehouse operations are a key enabler to a successful supply chain implementation and the starting point for a transformation. It is critical to manage safety, quality and efficiency. Learn to leverage the lean supply chain modifications to improve customer responsiveness and reduce operating costs and in doing so contributing to a supply chain that creates a competitive advantage for a company. To accomplish this goal, we must bring lean principles into the warehouse and distribution center.
Who Should Attend
Supply chain professionals, logistics professionals, material managers, production control managers, transportation managers, warehousing managers and purchasing managers
Upon completion of this course, you will be able to:
Benefits:
Abstract:
Traditionally, there are two pillars of science: theory and experimentation. These two inform one another and lead scientists to make educated guesses and decisions toward advancing science. More recently, the driving force behind scientific advancement has not just focused on how much information can be learned, but how quickly. Additionally, experimental data can be costly and difficult to obtain. With these motivations in mind, the field of experimental design aims to maximize the information gained from as few experimental data points as possible. Computation has emerged as a third pillar of science to complement the traditional two and has been used to facilitate optimal experimental design.
Sandia’s Z machine is the world’s most powerful and efficient laboratory radiation source. Z experiments often exhibit large current losses, so a principal uncertainty is how effectively current can be delivered. Power flow simulations are very intensive, making them infeasible to use in critical design and optimization studies. Developing a consistent picture of how losses develop and evolve would improve understanding of present-day experiments and better constrain circuit model representations, providing a basis for quantifying uncertainties in circuit models applied to Z and improve confidence in predictions of target performance. This presentation details the implementation of a Bayesian optimization study to maximize the information gained from Z experimental data and design.
* SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525
Bio:
Kathryn Maupin is a Principal Member of the Technical Staff at Sandia National Laboratories. Motivated by a passion for transforming uncertainty into actionable insights, Kathryn leverages her extensive expertise in model validation, model form error quantification, and Bayesian analyses to drive innovative solutions that enhance research outcomes.
Kathryn earned her PhD in Computational Science, Engineering, and Mathematics, along with her M.S. in Computational and Applied Mathematics, both from The University of Texas at Austin. Her fascination with mathematical modeling began at the University of California, San Diego, where she completed her B.A. in Applied Mathematics.
When she is not immersed in data and algorithms, Kathryn enjoys the chaos of family life with her three children and three dogs. Looking ahead, Kathryn aspires to continue pushing the boundaries of computational science while encouraging others to confront ubiquitous uncertainty in their work.
]]>Traditionally, there are two pillars of science: theory and experimentation. These two inform one another and lead scientists to make educated guesses and decisions toward advancing science. More recently, the driving force behind scientific advancement has not just focused on how much information can be learned, but how quickly. Additionally, experimental data can be costly and difficult to obtain. With these motivations in mind, the field of experimental design aims to maximize the information gained from as few experimental data points as possible. Computation has emerged as a third pillar of science to complement the traditional two and has been used to facilitate optimal experimental design.
]]>
Dealing with Ambiguity in Humanitarian Decision-Making
Abstract:
One of the major challenges humanitarian organizations face in response planning is managing the inherent ambiguity and uncertainty of disaster situations. In post-disaster contexts, information from various sources (assessing both the needs of affected populations and the extent of damage in the impacted area) often contains missing elements and inconsistencies, which can hinder effective decision-making. In this talk, I will present a new methodological framework that combines graph clustering and stochastic optimization to support humanitarian decision-makers in analyzing the implications of divergent estimates from multiple data sources and in efficiently integrating these estimates into the decision-making process. The usefulness of the proposed approach is demonstrated through a realistic case study on shelter location planning for internally displaced people (IDPs) in a conflict setting, specifically the Syrian civil war. We use needs assessment data from two reliable sources toestimate shelter requirements in Idleb, a district of Syria. Our case study shows that the framework enables decision-makers to assess the degree of ambiguity in the data and the level of consensus across sources, ultimately supporting better-informed decisions and more effective planning for the delivery of humanitarian aid.
Bio:
Walter Rei is a Professor of Operations Research in the Department of Analytics, Operations, and Information Technologies at the École des Sciences de la Gestion, Université du Québec à Montréal, Canada. He currently holds the Canada Research Chair in Stochastic Optimization of Transport and Logistics Systems and is also a member of the Interuniversity Research Centre on Enterprise Networks, Logistics and Transportation (CIRRELT). His research focuses on the development of efficient solution methodologies for integer programs and combinatorial optimization models relevant to transportation and logistics problems involving uncertainty.
]]>
]]>
The Search for Parking for Commercial Last-Mile Delivery in Urban Environments (or should they?)
Abstract:
Parking is a time-consuming, and thus expensive, part of last-mile delivery in urban environments. To build insights into driver parking behavior, we introduce the Stochastic Parking Problem (SPP) to model the search process for parking where delivery drivers may choose to park at available parking spots or double park at unavailable parking spots at the risk of receiving a fine. To this end, the Stochastic Parking Problem (SPP) models the parking search process as a Markov Decision Process. We characterize the structure of the optimal parking policy for the SPP based on the probability that each parking spot is available and the expected cost of double parking. Further, we provide a polynomial-time algorithm to find the optimal policy for the SPP. We utilize the optimal policy for the SPP to derive managerial insights regarding how drivers should approach the parking search process. In doing so, we identify sufficient enforcement levels to eliminate double parking from optimal parking decisions for last-mile delivery drivers.
Bio:
Ann Melissa Campbell is the Clement T. and Sylvia H. Hanson Family Chair in Manufacturing Productivity in the Department of Business Analytics at the Henry B. Tippie College of Business. Her research focuses on freight transportation, especially on problems related to new and emerging business models. She is a recipient of the NSF CAREER Award and serves as an Area Editor for Transportation Science. As department chair, she led the department’s efforts to win the 2021 INFORMS UPS George D. Smith Prize for excellence in analytics education. Since 2022, she has chaired the annual FutureBAProf workshop focused on educating PhD students and postdocs about academic careers in business schools.
Read the full story here: https://www.isye.gatech.edu/news/turn-learning-action-industrial-engineering-community-research-projects
]]>Venue/Location
Georgia Tech Exhibition Hall - Home Park Room (2nd floor)
Georgia Institute of Technology
460 4th St NW, Atlanta, GA 30332
Join us for an insightful seminar where Dr. Parrish will delve into the complex interplay between tariffs, inflation, interest rates, and other economic factors. The seminar will explore how these elements influence not only the economy of Georgia but also the broader American and global economies. Dr. Parrish will provide an analysis of current economic trends, shedding light on the uncertainties and challenges faced by businesses and policymakers in today's interconnected world. Attendees will gain a deeper understanding of the economic forces at play and their potential impacts on trade, investment, and economic growth.
Jerry D. Parrish, Ph.D. is the Chief Economist at the Metro Atlanta Chamber. He previously served as the Director of State and Local Policy Analysis at the Florida Institute of Government at Florida State University, where he also spent twelve years on the faculty of the Masters of Applied Economics Program.
Dr. Parrish has held several prominent positions, including Chief Economist and Director of Research for the Florida Chamber Foundation, Chief Economist and Director of the Center for Competitive Florida at Florida TaxWatch, and Associate Director of the Center for Economic Forecasting & Analysis (CEFA) at Florida State University. His extensive experience in the private sector includes management roles at international manufacturing companies.
Jerry holds a B.S. in Agricultural Business and Economics from Auburn University, an M.B.A. from Bellarmine University, an M.S. in Economics from the University of North Carolina at Charlotte, and a Ph.D. in Economics from Auburn University.
For questions, please email event@scl.gatech.edu.
Analysis of the Genealogy Process in Forensic Investigative Genetic Genealogy
ABSTRACT:
The genealogy process is typically the most time-consuming part of -- and a limiting factor in the success of -- forensic investigative genetic genealogy, which is a new approach to solving violent crimes and identifying human remains. We formulate a stochastic dynamic program that -- given the list of matches and their genetic distances to the unknown target -- chooses the best decision at each point in time: which match to investigate (i.e., find its ancestors), which ancestors of these matches to descend from (i.e., find its descendants), or whether to terminate the investigation. The objective is to maximize the probability of finding the target minus a cost on the expected size of the final family tree. We estimate the parameters of our model using data from 17 cases (eight solved, nine unsolved) from the DNA Doe Project. We assess the Proposed Strategy using simulated versions of the 17 DNA Doe Project cases, and compare it to a Benchmark Strategy that ranks matches by their genetic distance to the target and only descends from known common ancestors between a pair of matches. The Proposed Strategy solves cases 25-fold faster than the Benchmark Strategy, and does so by aggressively descending from a set of potential most recent common ancestors between the target and a match even when this set has a low probability of containing the correct most recent common ancestor. This work has been used to solve several stalled cold cases.
BIO:
Lawrence M. Wein is the Jeffrey S. Skoll Professor of Management Science at the Graduate School of Business, Stanford University. He received a Ph.D. in Operations Research at Stanford University in 1988 and was a professor at MIT's Sloan School of Management from 1988 to 2002. His research interests are in operations management and public health. He was Editor-in-Chief of Operations Research from 2000 to 2005. He has been awarded a Presidential Young Investigator Award, the Erlang Prize, the Koopman Prize, the INFORMS Expository Writing Award, the Philip McCord Morse Lectureship, the INFORMS President’s Award, the Frederick W. Lanchester Prize, the George E. Kimball Medal, a best paper award from Risk Analysis, and two notable paper awards from the Journal of Forensic Sciences. He is an INFORMS Fellow, a M&SOM Fellow and a member of the National Academy of Engineering.
]]>]]>
Linear regression using Hilbert-space valued covariates with unknown reproducing kernel
Abstract:
In this talk we present a new method of linear regression using Hilbert-space valued covariates with unknown reproducing kernels. We develop a computationally efficient approach to estimation and derive asymptotic theory for the regression parameter estimates under mild assumptions. We demonstrate the approach in simulation studies as well as in a data analyses using two- and three-dimensional brain images as predictors. The is work is a collaboration with Xinyi Li and Margaret Hoch.
Bio:
Michael R. Kosorok, PhD, is the W.R. Kenan, Jr. Distinguished Professor of Biostatistics, Professor of Statistics and Operations Research, and Director of the Center for Artificial Intelligence and Public Health at the University of North Carolina at Chapel Hill. His interests include biostatistics, artificial intelligence, empirical processes, and precision health. He is a fellow of the ASA, IMS and AAAS, and is past-president of IMS.
]]>March 27, 2025
12pm ET | Complimentary Lunch (first 75 attendees)
1-2pm ET | Seminar
Callaway Manufacturing Research Building (Auditorium 101)
Ishai Menache will describe how LLMs and optimization techniques are leveraged in Microsoft’s cloud supply chain to improve efficiency, provide insights to planners, and answer "what-if" questions. More broadly, Menache will discuss ongoing research and future directions in the intersection of operations research and AI.
Ishai Menache is a Partner Researcher & Manager of the Machine Learning and Optimization group at Microsoft Research. With a PhD in Electrical Engineering from the Technion and postdoctoral experience at MIT, he has been with Microsoft since 2011. His research focuses on developing large-scale optimization frameworks that leverage machine learning and Generative AI technologies to improve supply chain operations.
Callaway Manufacturing Research Center (MARC)
Main Auditorium (1st Floor)
Georgia Institute of Technology
813 Ferst Drive NW
Atlanta, GA 30332
GPS Coordinates: 33.777873, -84.401353
Please note that we will be serving a free box lunch from 12-1pm (for the first 75 attendees) outside of the auditorium before the lecture at 1pm ET.
For questions, please email event@scl.gatech.edu.
💼 "Women in Consulting" Panel (PwC ISyE Alumnae)
📅 Tuesday, March 11 | 🕚 11:00 AM - 12:00 PM
📍 ISyE Main Atrium
🔹Presented by: PwC ISyE Alumnae
RSVP for the PwC seminar.
🎢"My Crazy IE Career Rollercoaster: Pivoting is What We Do?"
📅 Tuesday, March 25 | 🕚 11:00 AM - 12:00 PM
📍 ISyE Main Atrium
🔹Presented by Dr. Fay Cobb Payton, Special Advisor on Innovation and Professor, Mathematics and Computer Science at Rutgers University and Partner Consultant
RSVP for Fay Cobb Payton’s seminar.
These events provide a fantastic opportunity to learn from and celebrate the achievements of women in ISyE. Mark your calendars and join us!
]]>Carol English, carol.english@isye.gatech.edu
]]>
Thursday, April 3, 2025 | 12-1pm ET
Gain an overview of performance management and decision-making approaches in the field of supply chain engineering and management with Immersive Performance Management for Decision Making (i-PeM4DeM). Learners will participate in an experimental, hands-on activity to turn concepts and technique into action. The course concludes with an exploration of both the evolution and future of supply chain engineering and management. This is presented within the context of the modern supply chain, offering a totally new immersive and leading-edge experience.
This course is designed for supply chain, operations, and logistics executives looking to develop a better understanding of formal management principles; data analysts and data scientists working in business intelligence and interested in the emerging concept of business sentience; and consultants and advisors developing new management approaches for their clients.
The (Surprising) Rate Optimality of Greedy Procedures for Large-Scale Ranking and Selection
Abstract:
Ranking and selection (R&S) aims to select the best alternative with the largest mean performance from a finite set of alternatives. Recently, considerable attention has turned toward the large-scale R&S problem which involves a large number of alternatives. Ideal large-scale R&S procedures should be sample optimal; that is, the total sample size required to deliver an asymptotically nonzero probability of correct selection (PCS) grows at the minimal order (linear order) in the number of alternatives, k. Surprisingly, we discover that the naïve greedy procedure, which keeps sampling the alternative with the largest running average, performs strikingly well and appears sample optimal. To understand this discovery, we develop a new boundary-crossing perspective and prove that the greedy procedure is sample optimal for the scenarios where the best mean maintains at least a positive constant away from all other means as k increases. We further show that the derived PCS lower bound is asymptotically tight for the slippage configuration of means with a common variance. For other scenarios, we consider the probability of good selection and find that the result depends on the growth behavior of the number of good alternatives: if it remains bounded as k increases, the sample optimality still holds; otherwise, the result may change. Moreover, we propose the explore-first greedy procedures by adding an exploration phase to the greedy procedure. The procedures are proven to be sample optimal and consistent under the same assumptions. Last, we numerically investigate the performance of our greedy procedures in solving large-scale R&S problems.
This is a joint work with Zaile Li at INSEAD and Weiwei Fan at Tongji University. The paper is available at https://doi.org/10.1287/mnsc.2023.00694.
Bio:
Jeff Hong received his bachelor’s degree from Tsinghua University and his Ph.D. from Northwestern University. He is currently a professor in the Department of Industrial and Systems Engineering at the University of Minnesota, Twin Cities. Previously, he held faculty positions at Fudan University, the City University of Hong Kong, and the Hong Kong University of Science and Technology. His research interests include stochastic simulation, stochastic optimization, risk management, and supply chain management. Jeff is the Simulation Department Editor of Naval Research Logistics and an Associate Editor for Management Science and ACM Transactions on Modeling and Computer Simulation. He served as the Simulation Area Editor for Operations Research from 2018 to 2023, and as President of the INFORMS Simulation Society from 2020 to 2022.
]]>This is a joint work with Zaile Li at INSEAD and Weiwei Fan at Tongji University. The paper is available at https://doi.org/10.1287/mnsc.2023.00694.
Smarter decisions for a secure world: opportunities and challenges for industrial engineering
Abstract:
Our nation has undergone substantial transformation in the quarter century since the events of September 11, 2001, an era characterized by the need for security and resilience. During this time, the industrial engineering and operations research (IE/OR) community rose to this challenge, making remarkable advances by addressing these vital needs. This seminar will explore these critical topics. Advancing IE/OR through societally relevant applications has been a central theme of Dr. Laura Albert’s academic research career. In this talk, she will discuss recent advances from her research group in the areas of cybersecurity and critical infrastructure protection. The talk will end with an outline of some of the “grand challenges” in IE/OR. Prof. Albert will introduce a roadmap for advancing IE/OR by addressing these challenges and ensuring that IE/OR remains at the forefront of solving global problems.
Bio:
Laura Albert is a Professor of Industrial & Systems Engineering at the University of Wisconsin-Madison, and she was also the 2023 President of the Institute for Operations Research and the Management Sciences (INFORMS). Professor Albert’s research interests are in the field of operations research and analytics with application to homeland security, emergency response, and public sector problems. She has been recognized with the American Association for the Advancement of Science (AAAS) Fellow Award, Institute of Industrial and Systems Engineers (IISE) Fellow Award, the INFORMS Impact Prize, a National Science Foundation CAREER award, and a Fulbright Award. She is also an engineering ambassador who regularly promotes operations research locally and nationally through media appearances and on her blog entitled “Punk Rock Operations Research.”
]]>Join us in celebrating the incredible legacy of Professor Alexander Shapiro at ShapiroFest!
This workshop, dedicated to honoring Professor Shapiro's 75th birthday, will take place on March 17-18, 2025, at the Georgia Institute of Technology. The event will feature distinguished speakers, research talks on modern stochastic optimization, engaging discussions, and reflections on Professor Shapiro's transformative influence on stochastic programming and optimization.
Professor Shapiro's pioneering work in stochastic programming has made significant contributions to the theoretical and methodological foundations of the field. His innovations include risk-averse optimization, distributionally robust Markov decision processes, duality theory, perturbation analysis, sample average approximation, and robust stochastic approximation. These advancements have expanded the scope and capabilities of stochastic programming, making it a critical tool in emerging fields such as machine learning and artificial intelligence.
As the A. Russell Chandler III Chair and Professor in the H. Milton Stewart School of Industrial and Systems Engineering at Georgia Tech, Professor Shapiro has received numerous prestigious awards, including the Khachiyan Prize, the Dantzig Prize, and the John von Neumann Theory Prize. His work has had a profound impact on the field, and ShapiroFest is a testament to his remarkable contributions.
We warmly invite researchers, practitioners, and students across all fields of science and engineering to join us in celebrating this milestone in Professor Shapiro's illustrious career.
For details about the program, speakers, and registration, please visit the workshop website. Registration for the workshop is free but required due to limited capacity.
Meeting demand in a timely and cost-effective manner is important both in public and private supply chains, and heavily depend on the design and operation of these supply chains. Demand is affected by ongoing factors such as local economy, infrastructure, and geographic location, as well as unexpected events such as natural or manmade disasters or other large-scale disruptions. Designing and operating responsive supply chains requires the consideration of uncertainty in timing, scope, scale, and understanding of various topics such as forecasting, distribution network design, and inventory management. This course will examine methods and models for making supply chain design and operational decisions and explore the significant value that is obtained through informed decision-making in advance of an unpredictable event or long-term strategy for meeting the need of customers and beneficiaries.
This course is designed for representatives from governmental or non-governmental organizations, private corporations, military, and foundations, including but not limited to senior executives overseeing administrative and operational functions of an organization, logistics and supply chain managers, program managers, directors of field operations, directors of emergency/disaster preparedness and response, and public health professionals.
This course is the first in a 3-part virtually synchronous professional education program. Register and pay for all three required Health and Humanitarian Supply Chain Management Certificate courses and receive a discount of $400 off per course. Enter coupon code SCL-HHS at checkout with the Georgia Tech Professional Education website..
Additionally, there are scholarships available for the certificate program. Apply at https://hhls.scl.gatech.edu/ by December 31, 2024.
Questions? Reach out to chhs@gatech.edu!
]]>Many Supply Chain decisions are concerned with the timely and efficient procurement, allocation, and distribution of resources (e.g. funds, supplies, volunteers, money, employees) through a supply chain network. This course will explore methodologies for “medium term” decision making including procurement and inventory policies, strategies for distribution and allocation of limited resources, and supply chain design.
This course is designed for representatives from governmental or non-governmental organizations, private corporations, military, and foundations, including but not limited to senior executives overseeing administrative and operational functions of an organization, logistics and supply chain managers, program managers, directors of field operations, directors of emergency/disaster preparedness and response, and public health professionals.
This course is the second in a 3-part virtually synchronous professional education program. Register and pay for all three required Health and Humanitarian Supply Chain Management Certificate courses and receive a discount of $400 off per course. Enter coupon code SCL-HHS at checkout with the Georgia Tech Professional Education website..
Additionally, there are scholarships available for the certificate program. Apply at https://hhls.scl.gatech.edu/ by December 31, 2024. .
Questions? Reach out to chhs@gatech.edu!
]]>Often the lack of cooperation and coordination between organizations or stakeholders lead to inefficiencies, despite having common goals. A systems view is needed to ensure appropriate use of scarce resources to meet the multiple, and often conflicting, short- and long-term goals from multiple constituents. This course will focus on conceptual and modeling skills to understand and effectively manage supply chains and operations from a systems perspective. Models will address system characteristics (e.g., demand dependencies) that drive system dynamics and policies to regulate performance. Course topics include methods for improving coordination and collaboration, addressing demand dependencies, and reliably measuring and evaluating system performance.
This course is designed for representatives from governmental or non-governmental organizations, private corporations, military, and foundations, including but not limited to senior executives overseeing administrative and operational functions of an organization, logistics and supply chain managers, program managers, directors of field operations, directors of emergency/disaster preparedness and response, and public health professionals.
This course is the first in a 3-part virtually synchronous professional education program. Register and pay for all three required Health and Humanitarian Supply Chain Management Certificate courses and receive a discount of $400 off per course. Enter coupon code SCL-HHS at checkout with the Georgia Tech Professional Education website..
Additionally, there are scholarships available for the certificate program. Apply at https://hhls.scl.gatech.edu/ by December 31, 2024.
Questions? Reach out to chhs@gatech.edu!
]]>This course provides a deep dive into the ways in which artificial intelligence (AI) optimizes supply chain efficiency. Participants will explore generative AI fundamentals, prompt engineering, and practical applications such as automated inventory systems, predictive maintenance, and route optimization. The course also covers ethical AI use, good and bad use of generative AI (GenAI), and rapidly emerging use cases. By the end, professionals will be skilled in applying AI to enhance supply chain processes and drive success in their organizations.
This course targets supply chain managers, data analysts, logistics professionals, procurement specialists, and business leaders aiming to harness GenAI for enhanced supply chain operations. It is ideal for those interested in GenAI-driven efficiency, strategic insights, and navigation of GenAI's role in transforming supply chain processes.
Modern Supply Chain Overview (MoSCOw) covers the fundamental areas of supply chain engineering and management along with the ways in which they could evolve over time. An experimental activity in virtual reality is included, which lets course participants apply what they learn and gain concrete implementation experience.
This course is intended for operational supply chain players (in both supply chain engineering and supply chain management) who wish to gain a better understanding of the systemic nature of supply chain ecosystems. This course is also aimed at consultants and advisors who wish to gain a more structured and comprehensive view of modern supply chains.
This course is the fourth in the 4-course Supply Chain Analytics Professional certificate program. It incorporates learning advanced analytics and mathematical optimization to find solutions for supply chain problems. You’ll learn how to use linear programming, mixed integer programming, and heuristics to conduct prescriptive analytics related to production processes, distribution networks, and routing. The course serves as a capstone for the program by culminating in a hackathon where you’ll design networks, inventory policies, and scenarios and then evaluate the outcomes via simulations.
The online version of the course is comprised of (4) half-day online instructor-led LIVE group webinars and pre-work (e.g. installing and testing software on your computer, testing connectivity with LMS and meeting software, etc.) to be completed before the first day of the course.
Experienced business professionals who perform or want to perform analytics to improve their supply chain management processes. They want to tackle strategic goals and to perform leading edge analytics projects that address the full complexity of supply chains.
Essentials of Negotiations and Stakeholder Influence level-sets the participants' understanding of negotiation influence and strengthens preparation, planning and execution activities involved with both simple and complex negotiations. The program includes industry techniques and tools for traditional supplier negotiations, as well as tips for internal cross-functional leadership. Participants walk away with a standard industry and customized individual experience which includes their personal Negotiation Style “DNA” to help them embrace their own natural tendencies and strengths. The program includes mock negotiations to reinforce techniques and tactics immediately in a “no judgement zone” environment.
This course is ideal for sourcing initiative leaders, project leaders, business unit leaders, operations managers, sales leaders and procurement & supply management-related professionals who are involved with supplier selection, contract development and supplier performance management.
Contracting and Legal Oversight provides participants with a holistic and integrated understanding of contract law, contract types, key industry standard contract terms, and contract structure to improve their confidence when creating or modifying contract documents. The program is geared to reinforce standards of excellence for professionals who are responsible for delivering contractual agreements and mitigating financial risk for their organization.
The online version of the course is comprised of (3) instructor-led LIVE group webinars, homework, and pre-work (e.g. installing and testing software on your computer, testing connectivity with Canvas LMS and BlueJeans meeting software, etc.) to be completed before the first day of the course.
This course is ideal for contract managers, procurement professionals, sourcing initiative leaders, project managers and all procurement & supply management-related professionals involved with bid contract development, contract execution or supplier performance management.
Category Management and Sourcing Leadership is designed to deepen participants' knowledge base of core activities in the procurement & supply management function. The program covers the sourcing process, specifications gathering, common bid package alternatives, cross-functional collaboration and supplier evaluation & selection. Participants will walk away ready to develop bid packages more thoroughly to help drive sourcing decisions for their organizations. This "hands on" delivery focuses on the professional serving as the main liaison between the buying organization and the selling organization in the company sourcing process.
This course is ideal for sourcing initiative leaders, procurement professionals, project managers, finance analyst, contract managers and all procurement & supply management-related professionals involved with bid package development, bid package analysis, negotiations preparation, contracting and supplier selection activity.
This course provides a deep dive into the ways in which artificial intelligence (AI) optimizes supply chain efficiency. Participants will explore generative AI fundamentals, prompt engineering, and practical applications such as automated inventory systems, predictive maintenance, and route optimization. The course also covers ethical AI use, good and bad use of generative AI (GenAI), and rapidly emerging use cases. By the end, professionals will be skilled in applying AI to enhance supply chain processes and drive success in their organizations.
This course targets supply chain managers, data analysts, logistics professionals, procurement specialists, and business leaders aiming to harness GenAI for enhanced supply chain operations. It is ideal for those interested in GenAI-driven efficiency, strategic insights, and navigation of GenAI's role in transforming supply chain processes.
This course is the first in the four-course Supply Chain Analytics Professional certificate program. It prepares you to apply leading-edge analytical methods and technology enablers across the supply chain. You’ll learn the dynamics of supply chains, the most relevant planning challenges, and the roles of different types of analytics. Next, you’ll learn about data cleansing, exploratory data analysis, and visualization. You’ll use Python and PowerBI to analyze the causes of underperformance and to build dashboards to visualize supply chain data. You will leave knowing how to gather, analyze, and prepare your data through descriptive analytics before you dig into deeper applications.
The online version of the course is comprised of (4) half-day instructor-led LIVE group webinars and pre-work (e.g. installing and testing software on your computer, testing connectivity with LMS and meeting software, etc.) to be completed before the first day of the course.
Experienced business professionals who perform or want to perform analytics to improve their supply chain management processes. They want to tackle strategic goals and to perform leading edge analytics projects that address the full complexity of supply chains.
Category Management and Sourcing Leadership is designed to deepen participants' knowledge base of core activities in the procurement & supply management function. The program covers the sourcing process, specifications gathering, common bid package alternatives, cross-functional collaboration and supplier evaluation & selection. Participants will walk away ready to develop bid packages more thoroughly to help drive sourcing decisions for their organizations. This "hands on" delivery focuses on the professional serving as the main liaison between the buying organization and the selling organization in the company sourcing process.
This course is ideal for sourcing initiative leaders, procurement professionals, project managers, finance analyst, contract managers and all procurement & supply management-related professionals involved with bid package development, bid package analysis, negotiations preparation, contracting and supplier selection activity.
Contracting and Legal Oversight provides participants with a holistic and integrated understanding of contract law, contract types, key industry standard contract terms, and contract structure to improve their confidence when creating or modifying contract documents. The program is geared to reinforce standards of excellence for professionals who are responsible for delivering contractual agreements and mitigating financial risk for their organization.
The online version of the course is comprised of (3) instructor-led LIVE group webinars, homework, and pre-work (e.g. installing and testing software on your computer, testing connectivity with Canvas LMS and BlueJeans meeting software, etc.) to be completed before the first day of the course.
This course is ideal for contract managers, procurement professionals, sourcing initiative leaders, project managers and all procurement & supply management-related professionals involved with bid contract development, contract execution or supplier performance management.
Essentials of Negotiations and Stakeholder Influence level-sets the participants' understanding of negotiation influence and strengthens preparation, planning and execution activities involved with both simple and complex negotiations. The program includes industry techniques and tools for traditional supplier negotiations, as well as tips for internal cross-functional leadership. Participants walk away with a standard industry and customized individual experience which includes their personal Negotiation Style “DNA” to help them embrace their own natural tendencies and strengths. The program includes mock negotiations to reinforce techniques and tactics immediately in a “no judgement zone” environment.
This course is ideal for sourcing initiative leaders, project leaders, business unit leaders, operations managers, sales leaders and procurement & supply management-related professionals who are involved with supplier selection, contract development and supplier performance management.
This course is the second in the four-course Supply Chain Analytics Professional certificate program. It emphasizes operational performance metrics to align supply chain management with strategic business goals. You’ll learn several statistics concepts (e.g. variance analysis, hypothesis testing, forecasting methods) along with inventory management models. You’ll use diagnostic analytics with PowerBI and Python to conduct demand and service profiling, undertake root cause analysis, and use time series forecasting in inventory management.
The online version of the course is comprised of (4) half-day online instructor-led LIVE group webinars and pre-work (e.g. installing and testing software on your computer, testing connectivity with LMS and meeting software, etc.) to be completed before the first day of the course.
Experienced business professionals who perform or want to perform analytics to improve their supply chain management processes. They want to tackle strategic goals and to perform leading edge analytics projects that address the full complexity of supply chains.
Alexander Shapiro is the A. Russell Chandler III Chair and Professor in the H. Milton Stewart School of Industrial and Systems Engineering at Georgia Tech. Dr. Shapiro’s research interests are focused on stochastic programming, risk analysis, simulation based optimization, and multivariate statistical analysis. In 2013 he was awarded Khachiyan Prize of INFORMS for lifetime achievements in optimization, and in 2018 he was a recipient of the Dantzig Prize awarded by the Mathematical Optimization Society and Society for Industrial and Applied Mathematics. In 2020 he was elected to the National Academy of Engineering. In 2021 he was a recipient of John von Neumann Theory Prize awarded by the Institute for Operations Research and the Management Sciences (INFORMS). Dr. Shapiro served on the editorial boards of a number of professional journals. He was an area editor (optimization) of Operations Research and the editor-in-chief of Mathematical Programming, Series A.
]]>The workshop will delve into the broader community implications of AI and empower local leaders to support this technological shift. Solutions to public-sector issues such as traffic, public safety, and overall government efficiency will be discussed. Registration includes all course materials and meals for the day.
The workshop is made possible through funding from Georgia Artificial Intelligence in Manufacturing (Georgia AIM). Both the Center for Economic Development Research (CEDR) and Georgia AIM are part of Georgia Tech’s Enterprise Innovation Institute (EI2).
]]>Learning with Local and Global Adversarial Corruptions
We study learning in an adversarial setting, where an epsilon fraction of samples from a distribution P are globally corrupted (arbitrarily modified), and the remaining perturbations have an average magnitude bounded by rho (local corruptions). With access to n such corrupted samples, we aim to develop a computationally efficient approach that achieves the optimal minimax excess risk. Our approach combines a data-driven cleaning module with a distributionally robust optimization (DRO) framework. We demonstrate that if the data cleaning module is minimax optimal with respect to the Wasserstein loss, solving an optimal transport-based DRO problem ensures a minimax optimal decision. We further provide tractable reformulations for both modules. Specifically, we introduce an optimal filtering algorithm to clean corrupted data by identifying and removing outliers. For the DRO module, we reformulate the problem as a two-player zero-sum game, deriving finite convex formulations. We show that the minimax theorem applies to this game, and Nash equilibria exist. Finally, we present a principled approach for constructing adversarial examples.
Soroosh Shafiee is an assistant professor in the School of Operations Research and Information Engineering at Cornell University. Before that, he held positions as a postdoctoral researcher at both the Tepper School of Business at Carnegie Mellon University and the Automatic Control Laboratory at ETH Zurich. He held a B.Sc. and M.Sc. degree in Electrical Engineering from the University of Tehran and a Ph.D. degree in Operations Research from EPFL. His primary research interests revolve around data-driven optimization, low-complexity decision-making and optimal transport.
We study learning in an adversarial setting, where an epsilon fraction of samples from a distribution P are globally corrupted (arbitrarily modified), and the remaining perturbations have an average magnitude bounded by rho (local corruptions). With access to n such corrupted samples, we aim to develop a computationally efficient approach that achieves the optimal minimax excess risk. Our approach combines a data-driven cleaning module with a distributionally robust optimization (DRO) framework. We demonstrate that if the data cleaning module is minimax optimal with respect to the Wasserstein loss, solving an optimal transport-based DRO problem ensures a minimax optimal decision. We further provide tractable reformulations for both modules. Specifically, we introduce an optimal filtering algorithm to clean corrupted data by identifying and removing outliers. For the DRO module, we reformulate the problem as a two-player zero-sum game, deriving finite convex formulations. We show that the minimax theorem applies to this game, and Nash equilibria exist. Finally, we present a principled approach for constructing adversarial examples.
]]>A novel association study framework powered by machine learning
Abstract:
Genome-wide association studies (GWASs) have been widely applied to discover genetic variants associated with a diverse array of traits. So far, almost all GWASs conducted in neuroimaging genetics are performed on image-derived quantitative features, which are univariate. In this study, we proposed and implemented a novel machine learning strategy for systematically identifying variants that lead to detectable discrepancies on the full-frame brain images. When applied to data collected by the Alzheimer’s Disease Neuroimaging Initiative (ADNI) consortium, we are able to identify novel variants that show strong association with brain phenotypes.
]]>
Genome-wide association studies (GWASs) have been widely applied to discover genetic variants associated with a diverse array of traits. So far, almost all GWASs conducted in neuroimaging genetics are performed on image-derived quantitative features, which are univariate. In this study, we proposed and implemented a novel machine learning strategy for systematically identifying variants that lead to detectable discrepancies on the full-frame brain images. When applied to data collected by the Alzheimer’s Disease Neuroimaging Initiative (ADNI) consortium, we are able to identify novel variants that show strong association with brain phenotypes.
]]>
The workshop will delve into the broader community implications of AI and empower local leaders to support this technological shift. Solutions to public-sector issues such as traffic, public safety, and overall government efficiency will be discussed. Registration includes all course materials and meals for the day.
The workshop is made possible through funding from Georgia Artificial Intelligence in Manufacturing (Georgia AIM). Both the Center for Economic Development Research (CEDR) and Georgia AIM are part of Georgia Tech’s Enterprise Innovation Institute (EI2).
]]>On Principal Component Regression in High Dimension
Principal component regression (PCR) is a classical two-step approach to linear regression, where one first reduces the data dimension by projecting onto its leading principal components, and then performs ordinary least squares regression. We study PCR in an asymptotic high-dimensional regression setting, where the number of data points is proportional to the dimension. Our main deliverables are asymptotically exact limiting formulas for the estimation and prediction risks, which depend in a nuanced way on the eigenvalues of the population covariance, the alignment between the population principal components and the true signal, and the number of selected components.
A key challenge in the high-dimensional regime is that the sample covariance matrix is an inconsistent estimate of its population counterpart, and thus sample principal components may fail to capture potential latent low-dimensional structure in the data. We demonstrate this point through several case studies, including that of a spiked covariance matrix. The analysis of (random design) linear regression in high dimension typically builds on powerful results from random matrix theory, such as the Marchenko–Pastur law and deterministic equivalents for the resolvent of a sample covariance matrix. However, these standard tools alone are not sufficient for analyzing the prediction risk of PCR. To that end, we leverage and develop somewhat less standard techniques, which, to our knowledge, have not seen wide use in the statistics literature to date: multi-resolvent traces and their associated eigenvector overlap measures.
Based on joint work with Alden Green (Stanford).
https://arxiv.org/abs/2405.11676
Elad Romanov is a postdoctoral researcher in the Department of Statistics, Stanford, where he is hosted by Prof. David Donoho. Prior to that, he completed his PhD in the School of Computer Science, the Hebrew University of Jerusalem, where he was advised by Profs. Or Ordentlich and Matan Gavish. His research interests broadly span high-dimensional statistics, information theory and signal processing, and the mathematics of data science.
]]>Principal component regression (PCR) is a classical two-step approach to linear regression, where one first reduces the data dimension by projecting onto its leading principal components, and then performs ordinary least squares regression. We study PCR in an asymptotic high-dimensional regression setting, where the number of data points is proportional to the dimension. Our main deliverables are asymptotically exact limiting formulas for the estimation and prediction risks, which depend in a nuanced way on the eigenvalues of the population covariance, the alignment between the population principal components and the true signal, and the number of selected components.
A key challenge in the high-dimensional regime is that the sample covariance matrix is an inconsistent estimate of its population counterpart, and thus sample principal components may fail to capture potential latent low-dimensional structure in the data. We demonstrate this point through several case studies, including that of a spiked covariance matrix. The analysis of (random design) linear regression in high dimension typically builds on powerful results from random matrix theory, such as the Marchenko–Pastur law and deterministic equivalents for the resolvent of a sample covariance matrix. However, these standard tools alone are not sufficient for analyzing the prediction risk of PCR. To that end, we leverage and develop somewhat less standard techniques, which, to our knowledge, have not seen wide use in the statistics literature to date: multi-resolvent traces and their associated eigenvector overlap measures.
Based on joint work with Alden Green (Stanford).
https://arxiv.org/abs/2405.11676
]]>
Yule’s “nonsense correlation”: Moments and density.
In 1926, G. Udny Yule considered the following problem: given two independent and identically distributed random walks independent from each other, what is the distribution of their empirical correlation coefficient? Yule empirically observed the distribution of this statistic to be heavily dispersed and frequently large in absolute value, leading him to call it “nonsense correlation.'' This unexpected finding led to his formulation of two concrete questions, each of which would remain open for more than ninety years: (i) Find (analytically) the variance of this empirical correlation coefficient and (ii): Find (analytically) the higher order moments and the density of this empirical correlation coefficient. After giving a brief overview of the solution to question (i) in Ernst et al. (The Annals of Statistics, 2017), we turn to the recent work of Ernst et al. (Bernoulli, 2025), which closed question (ii) by explicitly calculating all moments of the empirical correlation coefficient (up to order 16). This leads, for the first time, to an approximation to the density of Yule's nonsense correlation. The methodology of Ernst et al. (2025) further enables explicit calculations of the moments of the empirical correlation coefficient when the two independent Wiener processes are replaced by two correlated Wiener processes, two independent Ornstein-Uhlenbeck processes, and two independent Brownian bridges. We also succeed in proving a Central Limit Theorem for the case of two independent Ornstein-Uhlenbeck processes. This shows that Yule's “nonsense correlation” is indeed not “nonsense” for stochastic processes which admit stationary distributions. The talk concludes with a discussion of some concrete applications of our work to the study of weather and climate extremes. The latter is part of our ongoing collaboration with the U.S. Office of Naval Research (2018-present).
Philip Ernst is Chair (and Full Professor) in Statistics and Royal Society Wolfson Fellow at Imperial College London. He was previously an Assistant Professor (2014-2018), an Associate Professor (2019-2022), and a Full Professor (2022-2023), all at Rice University’s Department of Statistics. His research lies at the interface of applied probability and operations research. His work has been funded by the U.S. Office of Naval Research (ONR), the U.S. Army Research Office (ARO), the National Science Foundation (NSF), The Royal Society, and The British Academy. Ernst is the recipient of numerous international and national research awards, including: a 2026 Institute of Mathematical Statistics (IMS) Medallion Award & Lecture, a 2023 Henri Lebesgue Chair, a 2023 British Academy/Wolfson Fellowship, a 2022 Committee of Presidents of Statistical Societies (COPSS) Emerging Leader Award, the 2020 (inaugural) INFORMS Donald P. Gaver, Jr. Early Career Award for Excellence in Operations Research, a 2018 U.S. Army Research Office (ARO) Young Investigator Award, and the 2018 Institute of Mathematical Statistics (IMS) Tweedie New Researcher Award. Ernst is also highly invested in teaching; he won seven teaching awards in the eight years he was employed at Rice University (including the George R. Brown Prize for Excellence in Teaching, Rice University’s most prestigious teaching award). He currently serves as an associate editor for six journals: Journal of Stochastic Analysis, Journal of the American Statistical Association: Theory and Methods, Mathematics of Operations Research, Statistics and Probability Letters, Stochastics, and The American Statistician. He is also an elected member of IMS Council (2024-2027).
]]>In 1926, G. Udny Yule considered the following problem: given two independent and identically distributed random walks independent from each other, what is the distribution of their empirical correlation coefficient? Yule empirically observed the distribution of this statistic to be heavily dispersed and frequently large in absolute value, leading him to call it “nonsense correlation.'' This unexpected finding led to his formulation of two concrete questions, each of which would remain open for more than ninety years: (i) Find (analytically) the variance of this empirical correlation coefficient and (ii): Find (analytically) the higher order moments and the density of this empirical correlation coefficient. After giving a brief overview of the solution to question (i) in Ernst et al. (The Annals of Statistics, 2017), we turn to the recent work of Ernst et al. (Bernoulli, 2025), which closed question (ii) by explicitly calculating all moments of the empirical correlation coefficient (up to order 16). This leads, for the first time, to an approximation to the density of Yule's nonsense correlation. The methodology of Ernst et al. (2025) further enables explicit calculations of the moments of the empirical correlation coefficient when the two independent Wiener processes are replaced by two correlated Wiener processes, two independent Ornstein-Uhlenbeck processes, and two independent Brownian bridges. We also succeed in proving a Central Limit Theorem for the case of two independent Ornstein-Uhlenbeck processes. This shows that Yule's “nonsense correlation” is indeed not “nonsense” for stochastic processes which admit stationary distributions. The talk concludes with a discussion of some concrete applications of our work to the study of weather and climate extremes. The latter is part of our ongoing collaboration with the U.S. Office of Naval Research (2018-present).
]]>Thursday, March 6, 2025 | 12-1pm ET
Abstract:
Problem definition: Human donor milk provides critical nutrition for millions of infants who are born preterm each year. Donor milk is collected, processed, and distributed by milk banks. The macronutrient content of donor milk is directly linked to infant brain development and can vary substantially across donations, which is why multiple donations are typically pooled together to create a final product. Approximately half of all milk banks in North America do not have the resources to measure the macronutrient content of donor milk, which means pooling is done heuristically. For these milk banks, an approach is needed to optimize pooling decisions. Methodology/results: We propose a data-driven framework combining machine learning and optimization to predict macronutrient content of donations and then optimally combine them in pools, respectively. In collaboration with our partner milk bank, we collect a data set of milk to train our predictive models. We rigorously simulate milk bank practices to fine-tune our optimization models and evaluate operational scenarios such as changes in donation habits during the COVID-19 pandemic. Finally, we conduct a year-long trial implementation, where we observe the current nurse-led pooling practices followed by our intervention. Pools created by our approach meet clinical macronutrient targets approximately 31% more often than the baseline, although taking 60% less recipe creation time. Managerial implications: This is the first paper in the broader blending literature that combines machine learning and optimization. We demonstrate that such pipelines are feasible to implement in a healthcare setting and can yield significant improvements over current practices. Our insights can guide practitioners in any application area seeking to implement machine learning and optimization-based decision support.
Bio:
Timothy Chan is the Associate Vice-President and Vice-Provost, Strategic Initiatives at the University of Toronto, the Canada Research Chair in Novel Optimization and Analytics in Health, a Professor in the department of Mechanical and Industrial Engineering, and a Senior Fellow of Massey College. He was previously Director of the Centre for Healthcare Engineering, Director of the Centre for Analytics and AI Engineering, and Associate Director, Research and Thematic Programming, of the Data Sciences Institute. His primary research interests are in operations research, optimization, and applied machine learning, with applications in healthcare, medicine, sustainability, and sports.
Professor Chan received his B.Sc. in Applied Mathematics from the University of British Columbia (2002), and his Ph.D. in Operations Research from the Massachusetts Institute of Technology (2007). Before coming to Toronto, he was an Associate in the Chicago office of McKinsey and Company (2007-2009), a global management consulting firm. During that time, he advised leading companies in the fields of medical device technology, travel and hospitality, telecommunications, and energy on issues of strategy, organization, technology and operations.
Professor Chan currently holds editorial roles in seven academic journals, including Operations Research, Management Science, and M&SOM. He has served in a variety of leadership and service roles at INFORMS and CORS, including as President of the INFORMS Health Application Society. He has over 120 publications in refereed journals, and is co-author of an upcoming book entitled “Introduction to Markov Decision Processes”. He has graduated over 50 graduate students and postdoctoral fellows, and takes great pride in cultivating a healthy, inclusive, and productive lab environment.
Professor Chan has received numerous awards and honours for his research, teaching and service. Recent highlights include the President’s Teaching Award from the University of Toronto in 2024, 1st place in the research paper competition at the MIT Sloan Sports Analytics Conference in 2024, the INFORMS Prize for Teaching OR/MS Practice in 2023, the Pierskalla Best Paper Award from INFORMS Health Applications Society in 2023, 1st place in the INFORMS Case Competition in 2022, and the CORS Eldon Gunn Service Award in 2022. His research has been featured by the CBC, CTV News, Global News, Reuters, CNN, the Globe and Mail, the Toronto Star, Boston Globe, ESPN, Canadian Business Magazine, and World Economic Forum.
]]>
Modern Sampling Paradigms: from Posterior Sampling to Generative AI
Sampling from a target distribution is a recurring theme in statistics and generative artificial intelligence (AI). In statistics, posterior sampling offers a flexible inferential framework, enabling uncertainty quantification, probabilistic prediction, as well as the estimation of intractable quantities. In generative AI, sampling aims to generate unseen instances that emulate a target population, such as the natural distributions of texts, images, and molecules.
In this talk, I will present my works on designing provably efficient sampling algorithms, addressing challenges in both statistics and generative AI. (1) In the first part, I will focus on posterior sampling for Bayes sparse regression. In general, such posteriors are high-dimensional and contain many modes, making them challenging to sample from. To address this, we develop a novel sampling algorithm based on decomposing the target posterior into a log-concave mixture of simple distributions, reducing sampling from a complex distribution to sampling from a tractable log-concave one. We establish provable guarantees for our method in a challenging regime that was previously intractable. (2) In the second part, I will describe a training-free acceleration method for diffusion models, which are deep generative models that underpin cutting-edge applications such as AlphaFold, DALL-E and Sora. Our approach is simple to implement, wraps around any pre-trained diffusion model, and comes with a provable convergence rate that strengthens prior theoretical results. We demonstrate the effectiveness of our method on several real-world image generation tasks.
Lastly, I will outline my vision for bridging the fields of statistics and generative AI, exploring how insights from one domain can drive progress in the other.
Yuchen Wu is a departmental postdoctoral researcher in the Department of Statistics and Data Science at the Wharton School, University of Pennsylvania. She earned her Ph.D. in 2023 from Stanford University, where she was advised by Professor Andrea Montanari. Her research lies broadly at the intersection of statistics and machine learning, featuring generative AI, high-dimensional statistics, Bayesian inference, algorithm design, and data-driven decision making.
]]>Sampling from a target distribution is a recurring theme in statistics and generative artificial intelligence (AI). In statistics, posterior sampling offers a flexible inferential framework, enabling uncertainty quantification, probabilistic prediction, as well as the estimation of intractable quantities. In generative AI, sampling aims to generate unseen instances that emulate a target population, such as the natural distributions of texts, images, and molecules.
In this talk, I will present my works on designing provably efficient sampling algorithms, addressing challenges in both statistics and generative AI. (1) In the first part, I will focus on posterior sampling for Bayes sparse regression. In general, such posteriors are high-dimensional and contain many modes, making them challenging to sample from. To address this, we develop a novel sampling algorithm based on decomposing the target posterior into a log-concave mixture of simple distributions, reducing sampling from a complex distribution to sampling from a tractable log-concave one. We establish provable guarantees for our method in a challenging regime that was previously intractable. (2) In the second part, I will describe a training-free acceleration method for diffusion models, which are deep generative models that underpin cutting-edge applications such as AlphaFold, DALL-E and Sora. Our approach is simple to implement, wraps around any pre-trained diffusion model, and comes with a provable convergence rate that strengthens prior theoretical results. We demonstrate the effectiveness of our method on several real-world image generation tasks.
Lastly, I will outline my vision for bridging the fields of statistics and generative AI, exploring how insights from one domain can drive progress in the other.
]]>Challenges and Opportunities in Assumption-free and Robust Inference
With the growing application of data science to complex high-stakes tasks, ensuring the reliability of statistical inference methods has become increasingly critical. This talk considers two key challenges to achieving this goal: model misspecification and data corruption, highlighting their associated difficulties and potential solutions. In the first part, we investigate the problem of distribution-free algorithm risk evaluation, uncovering fundamental limitations for answering these questions with limited amounts of data. To navigate the challenge, we will also discuss how incorporating an assumption about algorithmic stability might help. The second part focuses on constructing robust confidence intervals in the presence of arbitrary data contamination. We show that when the proportion of contamination is unknown, uncertainty quantification incurs a substantial cost, resulting in optimal robust confidence intervals that must be significantly wider.
Yuetian Luo is a postdoctoral scholar in the Data Science Institute at the University of Chicago advised by Professor Rina Foygel Barber. He received his Ph.D. in Statistics from the University of Wisconsin-Madison under the supervision of Professor Anru Zhang. His research interests lie broadly in distribution-free inference, computational complexity of statistical inference, tensor learning, robust statistics, and non-convex optimization.
]]>With the growing application of data science to complex high-stakes tasks, ensuring the reliability of statistical inference methods has become increasingly critical. This talk considers two key challenges to achieving this goal: model misspecification and data corruption, highlighting their associated difficulties and potential solutions. In the first part, we investigate the problem of distribution-free algorithm risk evaluation, uncovering fundamental limitations for answering these questions with limited amounts of data. To navigate the challenge, we will also discuss how incorporating an assumption about algorithmic stability might help. The second part focuses on constructing robust confidence intervals in the presence of arbitrary data contamination. We show that when the proportion of contamination is unknown, uncertainty quantification incurs a substantial cost, resulting in optimal robust confidence intervals that must be significantly wider.
]]>Simple menus in robust screening
This talk investigates the design and effectiveness of simple selling mechanisms when a seller has only partial information about a buyer’s valuation distribution, obtained through market research or price experimentation. While robust screening offers stronger guarantees compared to deterministic pricing, it often involves complex menus with infinitely many options, posing implementation challenges. Our research introduces simple mechanisms with finite menus that balance performance guarantees with practical implementation. Using a unified framework for various ambiguity sets—including support, mean, and quantile—we derive optimal mechanisms and performance ratios for different menu sizes. Our findings reveal that modest menu sizes can closely approximate the benefits of optimal infinite-menu mechanisms. Remarkably, even a two-option menu significantly outperforms deterministic pricing.
We extend our results to multi-item mechanism design, where optimal mechanisms are complicated even with full knowledge of buyers’ valuation distributions. To address this challenge, we propose “semi-separable mechanisms,” where each item's allocation and payment rules depend only on its valuation and joint distributional information, but not on the valuations of other items. We prove that semi-separable mechanisms achieve the optimal performance ratio among all incentive-compatible and individually rational mechanisms when only marginal support information is available. Additionally, our framework accommodates settings where sellers possess aggregate valuation information for product bundles, further enhancing its practical applicability.
Shixin Wang is an Assistant Professor in the Department of Decisions, Operations and Technology at The Chinese University of Hong Kong. Before joining CUHK, she earned her Ph.D. in Operations Management from NYU Stern School of Business and a bachelor’s degree in Industrial Engineering from Tsinghua University. Her research focuses on developing simple, robust pricing policies in revenue management and designing sparse, reliable networks for supply chain and service systems. Her work has been recognized as a finalist in the INFORMS JFIG Paper Competition and the INFORMS Service Science Best Cluster Paper Award. Her research has been supported by funding from Hong Kong Research Grants Council (RGC).
]]>This talk investigates the design and effectiveness of simple selling mechanisms when a seller has only partial information about a buyer’s valuation distribution, obtained through market research or price experimentation. While robust screening offers stronger guarantees compared to deterministic pricing, it often involves complex menus with infinitely many options, posing implementation challenges. Our research introduces simple mechanisms with finite menus that balance performance guarantees with practical implementation. Using a unified framework for various ambiguity sets—including support, mean, and quantile—we derive optimal mechanisms and performance ratios for different menu sizes. Our findings reveal that modest menu sizes can closely approximate the benefits of optimal infinite-menu mechanisms. Remarkably, even a two-option menu significantly outperforms deterministic pricing.
We extend our results to multi-item mechanism design, where optimal mechanisms are complicated even with full knowledge of buyers’ valuation distributions. To address this challenge, we propose “semi-separable mechanisms,” where each item's allocation and payment rules depend only on its valuation and joint distributional information, but not on the valuations of other items. We prove that semi-separable mechanisms achieve the optimal performance ratio among all incentive-compatible and individually rational mechanisms when only marginal support information is available. Additionally, our framework accommodates settings where sellers possess aggregate valuation information for product bundles, further enhancing its practical applicability.
]]>Cooperation and the Design of Public Goods
Public transit systems face two, often conflicting design goals: ridership and coverage. The ridership goal involves serving as many people as possible, typically in high-density urban centers. Conversely, the coverage goal treats transit as a social service and measures its success by how good a service it provides to those who badly need it, including those in low-density suburban areas. And while the social significance of the coverage goal has only grown after decades of increasingly suburbanized poverty, a transit agency that takes this too far is likely to face fierce pressure from those who contribute to its resources (e.g., in the form of taxes and fare collection) but do not benefit from such a service plan. This tension speaks to a fundamental difficulty with designing public goods, including but not limited to transit, that take on a social service mission while trying to maintain broad popular support.
In this talk, I will approach the design of public goods from the perspective of cooperative game theory. I will introduce non-transferable utility (NTU) linear production (LP) games, which combine the essential game-theoretic elements of public goods with the modeling flexibility of linear programming. I will show that under mild and interpretable conditions, designs that maintain popular support are possible. However, this result is existential: I will show that testing whether a particular design maintains popular support is co-NP-complete. I will also demonstrate how, while one can in principle write a mixed-integer linear programming formulation for the set of popular designs, this approach is vastly impractical even for simple instances, and that natural approaches to obtain a polyhedral relaxation through cutting plane methods can be insufficient. This motivates further research on optimizing over this complicated yet well-structured set. Lastly, I will tie this theory back to transit with a data-driven implementation that illustrates the impact of maintaining popular support on the distribution of quality of service for coverage-oriented transit designs.
Juan Carlos Martínez Mori is a Schmidt Science Fellow and a President’s Postdoctoral Fellow with the H. Milton Stewart School of Industrial and Systems Engineering at the Georgia Institute of Technology. Prior to his current appointment, he was a Postdoctoral Fellow at the Simons Laufer Mathematical Sciences Institute (SLMath, formerly MSRI) as part of their thematic program on Algorithms, Fairness, and Equity. He earned his PhD in Applied Mathematics from Cornell University in 2023 and his BSc in Civil Engineering and minor in Computer Science from the University of Illinois at Urbana-Champaign in 2017.
His primary research interests span transportation, optimization, and game theory, with additional prior work in enumerative combinatorics and sports analytics. As a frequent transit rider, he is interested in mathematics that support more accessible and convenient public transportation.
]]>Public transit systems face two, often conflicting design goals: ridership and coverage. The ridership goal involves serving as many people as possible, typically in high-density urban centers. Conversely, the coverage goal treats transit as a social service and measures its success by how good a service it provides to those who badly need it, including those in low-density suburban areas. And while the social significance of the coverage goal has only grown after decades of increasingly suburbanized poverty, a transit agency that takes this too far is likely to face fierce pressure from those who contribute to its resources (e.g., in the form of taxes and fare collection) but do not benefit from such a service plan. This tension speaks to a fundamental difficulty with designing public goods, including but not limited to transit, that take on a social service mission while trying to maintain broad popular support.
In this talk, I will approach the design of public goods from the perspective of cooperative game theory. I will introduce non-transferable utility (NTU) linear production (LP) games, which combine the essential game-theoretic elements of public goods with the modeling flexibility of linear programming. I will show that under mild and interpretable conditions, designs that maintain popular support are possible. However, this result is existential: I will show that testing whether a particular design maintains popular support is co-NP-complete. I will also demonstrate how, while one can in principle write a mixed-integer linear programming formulation for the set of popular designs, this approach is vastly impractical even for simple instances, and that natural approaches to obtain a polyhedral relaxation through cutting plane methods can be insufficient. This motivates further research on optimizing over this complicated yet well-structured set. Lastly, I will tie this theory back to transit with a data-driven implementation that illustrates the impact of maintaining popular support on the distribution of quality of service for coverage-oriented transit designs.
]]>High-dimensional Clustering via A Latent Transformation Mixture Model
Cluster analysis is a fundamental task in machine learning. Several clustering algorithms have been extended to handle high-dimensional data by incorporating a sparsity constraint in the estimation of a mixture of Gaussian models. Though it makes some neat theoretical analysis possible, this type of approach is arguably restrictive for many applications. In this work we propose a novel latent transformation mixture model for clustering. The use of unspecified transformation makes the model much more flexible than the classical model-based clustering. Under the assumption that the optimal clustering admits a sparsity structure, we develop a new clustering algorithm named CESME for high-dimensional clustering. We offer a comprehensive analysis of CESME including identifiability, initialization, algorithmic convergence, and statistical guarantees on clustering. Extensive numerical study and real data analysis show that CESME outperforms the existing high-dimensional clustering algorithms in the literature.
Dr. Hui Zou is currently the Dr. Lynn Y.S. Lin Professor at the University of Minnesota. He earned his Ph.D. in Statistics from Stanford University in 2005. His primary research interests include statistical learning, high-dimensional models, statistical computing, and the application of modern statistical methods in business, health, and engineering. Dr. Zou is an elected fellow of the American Association for the Advancement of Science (AAAS), the Institute of Mathematical Statistics (IMS), and the American Statistical Association (ASA). He has published over 100 research articles, many of which are highly cited, including three that have been listed among the most-cited papers of all time in JRSSB, JASA, and JCGS. One of his papers in the Annals of Statistics was selected as the Best Paper in Applied Mathematics at the 8th International Congress of Chinese Mathematicians (ICCM 2019). Dr. Zou has also mentored 15 PhD students, two of whom received the COPSS Leadership Academy Award for Emerging Leaders in Statistics.
]]>Cluster analysis is a fundamental task in machine learning. Several clustering algorithms have been extended to handle high-dimensional data by incorporating a sparsity constraint in the estimation of a mixture of Gaussian models. Though it makes some neat theoretical analysis possible, this type of approach is arguably restrictive for many applications. In this work we propose a novel latent transformation mixture model for clustering. The use of unspecified transformation makes the model much more flexible than the classical model-based clustering. Under the assumption that the optimal clustering admits a sparsity structure, we develop a new clustering algorithm named CESME for high-dimensional clustering. We offer a comprehensive analysis of CESME including identifiability, initialization, algorithmic convergence, and statistical guarantees on clustering. Extensive numerical study and real data analysis show that CESME outperforms the existing high-dimensional clustering algorithms in the literature.
]]>Covariate adjustment in randomized experiments with missing outcomes and covariates
Covariate adjustment can improve precision in analyzing randomized experiments. With fully observed data, regression adjustment and inverse probability weighting are asymptotically equivalent in improving efficiency over unadjusted analysis. When some outcomes are missing, we consider combining these two adjustment methods with inverse probability of observation weighting for handling missing outcomes, and show that the equivalence between the two methods breaks down. Regression adjustment no longer ensures efficiency gain over unadjusted analysis unless the true outcome model is linear in covariates or the outcomes are missing completely at random. Propensity score weighting, in contrast, still guarantees efficiency over unadjusted analysis, and including more covariates in adjustment never harms asymptotic efficiency. Moreover, we establish the value of using partially observed covariates to secure additional efficiency by the missingness indicator method, which imputes all missing covariates by zero and uses the union of the completed covariates and corresponding missingness indicators as the new, fully observed covariates. Based on these findings, we recommend using regression adjustment in combination with the missingness indicator method if the linear outcome model or missing complete at random assumption is plausible and using propensity score weighting with the missingness indicator method otherwise.
Fan Li is a professor in the Department of Statistical Science at Duke University, with a secondary appointment at Biostatistics and Bioinformatics. Her primary research interest in causal inference, and have been working at the intersection between causal inference, machine learning, and population health. She also works on Bayesian analysis and missing data problems. She is the editor for Social Science, Biostatistics and Policy of the Annals of Applied Statistics, and an elected fellow of the American Statistical Association and the Institute of Mathematical Statistics (IMS).
]]>Covariate adjustment can improve precision in analyzing randomized experiments. With fully observed data, regression adjustment and inverse probability weighting are asymptotically equivalent in improving efficiency over unadjusted analysis. When some outcomes are missing, we consider combining these two adjustment methods with inverse probability of observation weighting for handling missing outcomes, and show that the equivalence between the two methods breaks down. Regression adjustment no longer ensures efficiency gain over unadjusted analysis unless the true outcome model is linear in covariates or the outcomes are missing completely at random. Propensity score weighting, in contrast, still guarantees efficiency over unadjusted analysis, and including more covariates in adjustment never harms asymptotic efficiency. Moreover, we establish the value of using partially observed covariates to secure additional efficiency by the missingness indicator method, which imputes all missing covariates by zero and uses the union of the completed covariates and corresponding missingness indicators as the new, fully observed covariates. Based on these findings, we recommend using regression adjustment in combination with the missingness indicator method if the linear outcome model or missing complete at random assumption is plausible and using propensity score weighting with the missingness indicator method ot herwise.
]]>Propagation of Shocks on Networks: Can Local Information Predict Survival?
Complex systems are often fragile, where minor disruptions can cascade into dramatic collapses. Epidemics serve as a prime example of this phenomenon, while the 2008 financial crisis highlights how a domino effect, originating from the small subprime mortgage sector, can trigger global repercussions. Similarly, a massive power outage in eastern North America was seemingly set off by a localized fault. The mathematical theory underlying these phenomena is both elegant and foundational, profoundly shaping the field of Network Science since its inception. In this talk, I will present a unifying mathematical model for network fragility and cascading dynamics and explore its deep connections to the theory of local-weak convergence, pioneered by Benjamini-Schramm and Aldous-Steele.
Dr. Souvik Dhara is an Assistant Professor at the Edwardson School of Industrial Engineering at Purdue University. He previously held positions as a Schramm Fellow, with a joint appointment at MIT Mathematics and Microsoft Research, a Simons-Berkeley Fellow at the Simons Institute for the Theory of Computing, UC Berkeley, and a Postdoctoral Research Associate at Brown University. Dr. Dhara earned his Ph.D. from the Department of Mathematics and Computer Science at Eindhoven University of Technology. In recognition of his doctoral work, he was awarded the Stieltjes Prize at the Dutch Mathematical Congress 2019. Dr. Dhara’s research lies at the intersection of applied probability and network science, with a primary focus on developing theoretical foundations for stochastic processes and algorithms on large-scale networks. His interests include cascades on networks, graph representation learning, and different notions of graph limits.
]]>Complex systems are often fragile, where minor disruptions can cascade into dramatic collapses. Epidemics serve as a prime example of this phenomenon, while the 2008 financial crisis highlights how a domino effect, originating from the small subprime mortgage sector, can trigger global repercussions. Similarly, a massive power outage in eastern North America was seemingly set off by a localized fault. The mathematical theory underlying these phenomena is both elegant and foundational, profoundly shaping the field of Network Science since its inception. In this talk, I will present a unifying mathematical model for network fragility and cascading dynamics and explore its deep connections to the theory of local-weak convergence, pioneered by Benjamini-Schramm and Aldous-Steele.
]]>Foundations of Private Optimization for Modern Machine Learning
How can we develop optimization algorithms for training machine learning models that preserve the privacy of individuals' training data? In this talk, I will present my work addressing this challenge through differential privacy (DP). Differential privacy offers a rigorous, quantifiable standard of privacy that limits potential leakage of training data. I will explore the fundamental limits of performance for differentially private optimization in modern machine learning, particularly within federated learning settings, and present scalable, efficient algorithms that achieve optimal accuracy under DP constraints. Additionally, these algorithms demonstrate strong empirical performance.
Andrew Lowy is a postdoctoral Research Associate at University of Wisconsin-Madison, advised by Stephen J. Wright. He received his PhD in Applied Math at University of Southern California under the supervision of Meisam Razaviyayn, where he was awarded the 2023 Center for Applied Mathematical Sciences (CAMS) Graduate Student Prize for outstanding research. His work has been published in leading venues in optimization, machine learning, and privacy, including SIOPT, NeurIPS, ICML, ICLR, ALT, AISTATS, ACM CCS, and the Journal of Privacy and Confidentiality. Prior to his doctoral studies, he completed his undergraduate studies at Princeton University and Columbia University.
Andrew’s research focuses on optimization for private, fair, and robust machine learning. His primary area of expertise is in differentially private optimization, with an emphasis on understanding fundamental limits and developing scalable algorithms that attain these limits.
]]>How can we develop optimization algorithms for training machine learning models that preserve the privacy of individuals' training data? In this talk, I will present my work addressing this challenge through differential privacy (DP). Differential privacy offers a rigorous, quantifiable standard of privacy that limits potential leakage of training data. I will explore the fundamental limits of performance for differentially private optimization in modern machine learning, particularly within federated learning settings, and present scalable, efficient algorithms that achieve optimal accuracy under DP constraints. Additionally, these algorithms demonstrate strong empirical performance.
]]>Surgical Human-Robot Collaborations: Transforming Training, Skills, and Safety
The introduction of surgical robotic systems in the operating room has changed the paradigm of how surgical team members physically and cognitively interact with both technology and one another. This shift has highlighted the need to reconsider nontechnical skills – task relevant cognitive and interpersonal skills among surgical team members – and to adapt surgical training accordingly. This presentation will highlight the impacts of robotic technology, both surgical systems and wearables, on skills training and clinician safety. It will also introduce potential interventions such as adaptive training programs and exoskeletons, aimed at improving surgical human-robot collaborations and team dynamics to promote surgical safety and performance.
Jackie Cha, PhD CPE is an Assistant Professor in the Department of Industrial Engineering and faculty with the Biomedical Data Science and Informatics Program at Clemson University and Faculty Scholar in the Clemson University School of Health Research. She obtained her PhD in Industrial Engineering from Purdue University, and MSE and BSE in Biomedical Engineering from the University of Michigan. Her research focuses on measuring physical human-robot interactions, particularly in healthcare (surgical) environments, to improve team performance, safety, and system efficiency. Her research has been funded by several sponsors such as the National Science Foundation (NSF) (including the NSF CAREER award), National Institutes of Health (NIH), and the Agency of Healthcare Research and Quality (AHRQ).
]]>The introduction of surgical robotic systems in the operating room has changed the paradigm of how surgical team members physically and cognitively interact with both technology and one another. This shift has highlighted the need to reconsider nontechnical skills – task relevant cognitive and interpersonal skills among surgical team members – and to adapt surgical training accordingly. This presentation will highlight the impacts of robotic technology, both surgical systems and wearables, on skills training and clinician safety. It will also introduce potential interventions such as adaptive training programs and exoskeletons, aimed at improving surgical human-robot collaborations and team dynamics to promote surgical safety and performance.
]]>Efficient Gradient Estimation for Overparameterized Stochastic Differential Equations
Overparameterized stochastic differential equation (SDE) models have achieved remarkable success in various complex environments, such as PDE-constrained optimization, stochastic control and reinforcement learning, financial engineering, neural SDEs, and generative AI. These models often feature system evolution coefficients that are parameterized by a high-dimensional vector θ in R^n, aiming to optimize expectations of the SDE, such as a value function, through stochastic gradient ascent. Consequently, designing efficient gradient estimators for which the computational complexity scales well with n is of significant interest. We introduce a novel unbiased stochastic gradient estimator--the generator gradient estimator--for which the computation time remains stable in n. In addition to establishing the validity of our methodology for general SDEs with jumps, we also perform numerical experiments testing our estimator in controlling a multi-class queue where the control policy is parameterized by high-dimensional neural networks. The results show a significant improvement in efficiency compared to previous methods: our estimator achieves near-constant computation times, increasingly outperforms its counterparts as n increases. These empirical findings highlight the potential of our proposed methodology for optimizing SDEs in contemporary applications.
This is a joint work with Jose Blanchet and Peter Glynn.
Shengbo Wang is a fifth-year Ph.D. candidate in Operations Research at Stanford University’s Department of Management Science and Engineering, co-advised by Prof. Peter Glynn and Prof. Jose Blanchet. His research interests span a broad spectrum within applied probability, including stochastic modeling, reinforcement learning, distributionally robust control, and simulation methods for machine learning. He focuses on developing tractable probabilistic models and designing algorithms for data-driven dynamic decision-making under uncertainty, specifically addressing reliability and scalability challenges in modern managerial and engineering applications. Prior to his PhD studies, he earned his B.S. in Operations Research and Information Engineering from Cornell University.
]]>Overparameterized stochastic differential equation (SDE) models have achieved remarkable success in various complex environments, such as PDE-constrained optimization, stochastic control and reinforcement learning, financial engineering, neural SDEs, and generative AI. These models often feature system evolution coefficients that are parameterized by a high-dimensional vector θ in R^n, aiming to optimize expectations of the SDE, such as a value function, through stochastic gradient ascent. Consequently, designing efficient gradient estimators for which the computational complexity scales well with n is of significant interest. We introduce a novel unbiased stochastic gradient estimator--the generator gradient estimator--for which the computation time remains stable in n. In addition to establishing the validity of our methodology for general SDEs with jumps, we also perform numerical experiments testing our estimator in controlling a multi-class queue where the control policy is parameterized by high-dimensional neural networks. The results show a significant improvement in efficiency compared to previous methods: our estimator achieves near-constant computation times, increasingly outperforms its counterparts as n increases. These empirical findings highlight the potential of our proposed methodology for optimizing SDEs in contemporary applications.
This is a joint work with Jose Blanchet and Peter Glynn.
]]>Advances in School District Design: Addressing Inequities and Planning for the Future
For decades, public school districts in the United States have faced complex decisions related to school district design. School desegregation in 1954 led to a series of operations research approaches to support these decisions. In this talk, we present a new modeling framework for the school district design problem, developed in partnership with a local school district, to facilitate an iterative community co-design process to address historic inequities in access to education and improve student assignment. This process led to recommendations and policy outcomes, including the school board’s approval of a new school and revised school attendance boundaries. At the core of our approach is a novel formulation that consolidates multiple assignment decisions, capturing their interactions through composite variables. The compact nature of this formulation makes it amenable to important extensions. Notably, we consider time-expanded versions of the model that allow districts to plan for multiple years. Importantly, these time-expanded district design models consider continuity in students’ educational experiences over time periods that may include the opening and closing of schools. This model can also incorporate uncertainty in future student enrollment over the planning horizon. Testing these new models is challenging given the lack of shareable data sets due to sensitive student information. To address this, we developed a method to create context-rich data sets for school operations models and methods relying only on publicly available data.
Aysu Ozel is a Ph.D. candidate in the Department of Industrial Engineering and Management Sciences at Northwestern University. Her dissertation research focuses on developing models and solution approaches to support decision-making for school district design. Her work has been recognized as the Winner of the 2023 INFORMS Doing Good with Good OR Student Paper Competition and a Finalist for the 2023 INFORMS DEI Best Student Paper Award. She is a Dissertation Year Fellow of Northwestern University Transportation Center and a Terminal Year Fellow of McCormick School of Engineering. Before her doctoral program, Aysu earned her M.S. and B.S. degrees in Industrial Engineering from Bilkent University.
]]>For decades, public school districts in the United States have faced complex decisions related to school district design. School desegregation in 1954 led to a series of operations research approaches to support these decisions. In this talk, we present a new modeling framework for the school district design problem, developed in partnership with a local school district, to facilitate an iterative community co-design process to address historic inequities in access to education and improve student assignment. This process led to recommendations and policy outcomes, including the school board’s approval of a new school and revised school attendance boundaries. At the core of our approach is a novel formulation that consolidates multiple assignment decisions, capturing their interactions through composite variables. The compact nature of this formulation makes it amenable to important extensions. Notably, we consider time-expanded versions of the model that allow districts to plan for multiple years. Importantly, these time-expanded district design models consider continuity in students’ educational experiences over time periods that may include the opening and closing of schools. This model can also incorporate uncertainty in future student enrollment over the planning horizon. Testing these new models is challenging given the lack of shareable data sets due to sensitive student information. To address this, we developed a method to create context-rich data sets for school operations models and methods relying only on publicly available data.
]]>Thursday, February 6, 2025 | 12-1pm ET
Thursday, January 2, 2025 | 12-1pm ET
Online Contention Resolution Schemes for the Matching Polytope of Graphs''
Online Contention Resolution Schemes (OCRS's) represent a modern tool for selecting a subset of elements, subject to resource constraints, when the elements are presented to the algorithm sequentially. OCRS's have led to some of the best-known competitive ratio guarantees for online resource allocation problems, with the added benefit of treating different online decisions---accept/reject, probing, pricing---in a unified manner. We analyze OCRS's for resource constraints defined by graph matchings, a fundamental structure in combinatorial optimization. We improve the state of the art both in terms of algorithmic guarantees and impossibility results. Our algorithms directly improve the best-known competitive ratios for online accept/reject, probing, and pricing problems on graphs. This includes the prophet matching problem for both edge and vertex arrival models, as well as for matching in a gig-economy. Our techniques are also relevant to more complicated resource constraints, and we attain new results for network revenue management and online assortment optimization.
Calum MacRury is a Postdoctoral Research Scholar in the Decision, Risk, and Operations Division at Columbia Business School, where he is supported by
an NSERC Postdoctoral Fellowship (PDF). He works on online algorithms, and more generally, decision making under uncertainty. He is particularly interested in stochastic optimization, including prophet inequalities and probing problems when the resource constraints are described by graph matchings. Calum received his PhD from the Department of Computer Science at the University of Toronto in 2023.
Online Contention Resolution Schemes (OCRS's) represent a modern tool for selecting a subset of elements, subject to resource constraints, when the elements are presented to the algorithm sequentially. OCRS's have led to some of the best-known competitive ratio guarantees for online resource allocation problems, with the added benefit of treating different online decisions---accept/reject, probing, pricing---in a unified manner. We analyze OCRS's for resource constraints defined by graph matchings, a fundamental structure in combinatorial optimization. We improve the state of the art both in terms of algorithmic guarantees and impossibility results. Our algorithms directly improve the best-known competitive ratios for online accept/reject, probing, and pricing problems on graphs. This includes the prophet matching problem for both edge and vertex arrival models, as well as for matching in a gig-economy. Our techniques are also relevant to more complicated resource constraints, and we attain new results for network revenue management and online assortment optimization.
]]>Integrative Artificial Intelligence for Healthcare
Developing integrated artificial intelligence frameworks is crucial for hospital operations and medical diagnostics. However, despite the opportunities, there remain challenges on 1) how to effectively learn, share, and combine information from diverse sources into a single setting and 2) how to design models that are practically implementable. In the first part, we propose and evaluate a unified Holistic AI in Medicine (HAIM) framework to facilitate the generation and testing of AI systems that integrate data from multiple modality sources, including tabular, time-series, language, and vision data. Our approach leverages recently developed open-source, pre-trained large models. We show that this framework can consistently and robustly produce models that outperform similar single-modality approaches across various healthcare tasks by 6–33%. In the second part, we introduce Multimodal Multitask Machine Learning for Healthcare (M3H), an explainable framework that consolidates learning for multiple tasks, including supervised binary/multiclass classification, regression, and unsupervised clustering in a single model to better exploit the interactions and dependencies among tasks. It introduces a novel attention mechanism inspired by healthcare considerations that designs a token-based query computation and scaling function that encourages self-exploitation and cross-exploration. The work, in addition, proposes a new explainability metric of the task space to better quantify the dynamics of task learning interdependencies and to automatically detect patterns among task relations. M3H encompasses a wide range of medical tasks and consistently outperforms traditional single-task models by an average of 11.6% across 44 medical tasks. The modular design of both frameworks ensures their generalizability in data processing, task definition, and rapid model prototyping. Combined, HAIM and M3H offer methodological and practical solutions to design integrative artificial intelligence to impact practice.
Yu Ma is a final year PhD student at the MIT Operations Research Center, where she is advised by Prof. Dimitris Bertsimas. Her research focuses on the use of AI methodologies to solve significant problems in healthcare service and policy making. Driven by a commitment to creating real-world impact, she has collaborated with six healthcare institutions and implemented three of her works in practice at Hartford Healthcare, the largest hospital system in Connecticut. In tackling these challenges, her works combine tools from machine learning, optimization, and analytics. She is recognized by the MIT School of Engineering as a Takeda Fellow. In the summer of 2023, she was an applied scientist in eBay’s Recommendation team. Prior to PhD, she obtained a B.A. degree from UC Berkeley in Applied Mathematics.
]]>
Developing integrated artificial intelligence frameworks is crucial for hospital operations and medical diagnostics. However, despite the opportunities, there remain challenges on 1) how to effectively learn, share, and combine information from diverse sources into a single setting and 2) how to design models that are practically implementable. In the first part, we propose and evaluate a unified Holistic AI in Medicine (HAIM) framework to facilitate the generation and testing of AI systems that integrate data from multiple modality sources, including tabular, time-series, language, and vision data. Our approach leverages recently developed open-source, pre-trained large models. We show that this framework can consistently and robustly produce models that outperform similar single-modality approaches across various healthcare tasks by 6–33%. In the second part, we introduce Multimodal Multitask Machine Learning for Healthcare (M3H), an explainable framework that consolidates learning for multiple tasks, including supervised binary/multiclass classification, regression, and unsupervised clustering in a single model to better exploit the interactions and dependencies among tasks. It introduces a novel attention mechanism inspired by healthcare considerations that designs a token-based query computation and scaling function that encourages self-exploitation and cross-exploration. The work, in addition, proposes a new explainability metric of the task space to better quantify the dynamics of task learning interdependencies and to automatically detect patterns among task relations. M3H encompasses a wide range of medical tasks and consistently outperforms traditional single-task models by an average of 11.6% across 44 medical tasks. The modular design of both frameworks ensures their generalizability in data processing, task definition, and rapid model prototyping. Combined, HAIM and M3H offer methodological and practical solutions to design integrative artificial intelligence to impact practice.
]]>
Two-Stage Stochastic Multi-Objective Linear Programming
Abstract:
We consider a two-stage stochastic multi-objective linear program (TSSMOLP) which is a natural multi-objective generalization of the well-studied two-stage stochastic linear program. The second-stage recourse decision is governed by an uncertain multi-objective linear program whose solution maps to an uncertain second-stage nondominated set. The TSSMOLP then comprises the objective function, which is the Minkowsi sum of a linear term plus the expected value of the second-stage nondominated set, and the constraints, which are linear. Since the second-stage nondominated set is a random set, its expected value is defined through the selection expectation. The global Pareto set is defined as the collection of nondominated points in the image space of the TSSMOLP. We discuss properties of TSSMOLPs and the multifunctions that arise therein, as well as the implications of these properties for the development of TSSMOLP solution methods. We illustrate the TSSMOLP and its properties through an example in disaster relief planning.
This work is joint work with Akshita Gupta, Edwardson School of Industrial Engineering, Purdue University.
Bio:
Susan R. Hunter is an associate professor in the Edwardson School of Industrial Engineering at Purdue University. Her research interests include theoretical and algorithmic aspects of stochastic optimization in the presence of multiple performance measures with emphasis on asymptotics, computation, and application. In 2016, she received an NSF CAREER Award to work on multi-objective simulation optimization; that is, multi-objective optimization in which the objective functions can only be observed with stochastic error as the output of a black-box Monte Carlo simulation oracle. Her published works have been recognized by the INFORMS Computing Society in 2011, by IISE Transactions in 2017, and by The Operational Research Society in 2021. She currently serves as Program Chair for the 2024 Winter Simulation Conference, on the Organizing Committee for SIAM Conference on Optimization (OP26), as Vice President / President Elect of the INFORMS Simulation Society, and as an associate editor for Operations Research, Journal of Optimization Theory and Applications, and Flexible Services and Manufacturing Journal.
]]>
Managing Tail Risk in Online Learning: When Safety Meets Efficiency
Online learning is a fast-growing research area in sequential decision-making. While previous work mostly focuses on achieving efficiency by minimizing regret expectation, controlling the regret tail risk to ensure safety is essential in applications such as revenue management, clinical trials, and financial investment, but has not been well studied. This work tries to provide a detailed characterization of regret distribution in online learning under the safety concern of managing tail risk.
In Part I, we aim to design policies that enjoy both optimal regret expectation and light-tailed regret distribution. We first find that any policy that obtains the optimal instance-dependent regret expectation could incur a heavy-tailed regret tail risk. We then design a novel policy that enjoys the optimal worst-case regret expectation and has the optimal worst-case regret tail risk with an optimal exponential decaying rate for any regret threshold. Numerical experiments show that our new policy design leads to similar efficiency and much better safety compared to celebrated policies. Our policy design also bears an interesting connection with Monte Carlo Tree Search (MCTS) used in AlphaGo. In Part II, we study the optimal trade-off between expectation and tail risk for regret distribution. We fully characterize the interplay among three desired properties for policy design: worst-case optimality, instance-dependent consistency, and light-tailed risk. Our results reveal several insights on how to design policies that balance efficiency and safety. All our results are extended to the stochastic linear bandit setting.
Feng Zhu is a 5th-year PhD student at MIT, advised by Prof. David Simchi-Levi. His research interests lie broadly in sequential decision-making under uncertainty, including online learning and online matching, with applications to online experimentation, supply chain management and revenue management. His research goal is to develop data-driven decision-making paradigms that ensure the safety & resiliency of modern operational systems, with a keen focus on managing (hidden) risk in various decision-making environments. Prior to MIT, he majored in Mathematics & Statistics and minored in Economics at Peking University.
]]>Online learning is a fast-growing research area in sequential decision-making. While previous work mostly focuses on achieving efficiency by minimizing regret expectation, controlling the regret tail risk to ensure safety is essential in applications such as revenue management, clinical trials, and financial investment, but has not been well studied. This work tries to provide a detailed characterization of regret distribution in online learning under the safety concern of managing tail risk.
In Part I, we aim to design policies that enjoy both optimal regret expectation and light-tailed regret distribution. We first find that any policy that obtains the optimal instance-dependent regret expectation could incur a heavy-tailed regret tail risk. We then design a novel policy that enjoys the optimal worst-case regret expectation and has the optimal worst-case regret tail risk with an optimal exponential decaying rate for any regret threshold. Numerical experiments show that our new policy design leads to similar efficiency and much better safety compared to celebrated policies. Our policy design also bears an interesting connection with Monte Carlo Tree Search (MCTS) used in AlphaGo. In Part II, we study the optimal trade-off between expectation and tail risk for regret distribution. We fully characterize the interplay among three desired properties for policy design: worst-case optimality, instance-dependent consistency, and light-tailed risk. Our results reveal several insights on how to design policies that balance efficiency and safety. All our results are extended to the stochastic linear bandit setting.
]]>Optimization under Uncertainty: Scheduling with Failover
Modern decision-making pipelines almost always rely on ML/AI tools to quantify uncertainty via probability distributions. This naturally leads to stochastic models, which we would like to optimize over to drive better decisions. I will discuss my work in developing fundamental algorithmic techniques for stochastic combinatorial optimization to drive algorithm design in both theory and practice.
I will introduce a novel resource allocation problem arising in cloud data centers: demands arrive online and need to be assigned to devices subject to capacity constraints to maximize utilization. Further, to be robust against failure, our assignment must remain feasible even in failover scenarios – when some device fails. These novel failover constraints introduce new trade-offs not present in classic assignment problems. We design asymptotically optimal online algorithms for this problem in both the worst- and average-case (where demands are drawn i.i.d. from an unknown distribution). Along the way, we will see a lesser-known probabilistic tool: stochastic monotone matchings. Preliminary experiments on real cloud workloads show the potential of our algorithms to generate more power-efficient allocations, which can save millions and significantly reduce the environmental impact of cloud computing.
Rudy is currently a postdoc at Carnegie Mellon, where he recently graduated with a PhD in the ACO (Algorithms, Combinatorics, and Optimization) program. He is broady interested in algorithm design - especially for discrete and stochastic optimization problems. His work draws on ideas from approximation algorithms, probability theory, and convex optimization to design improved algorithms and general-purpose technical tools for fundamental optimization problems. On the application side, he is interested in realizing the potential impact of theoretical algorithms in practice in areas such as data center scheduling and naval logistics.
]]>Modern decision-making pipelines almost always rely on ML/AI tools to quantify uncertainty via probability distributions. This naturally leads to stochastic models, which we would like to optimize over to drive better decisions. I will discuss my work in developing fundamental algorithmic techniques for stochastic combinatorial optimization to drive algorithm design in both theory and practice.
I will introduce a novel resource allocation problem arising in cloud data centers: demands arrive online and need to be assigned to devices subject to capacity constraints to maximize utilization. Further, to be robust against failure, our assignment must remain feasible even in failover scenarios – when some device fails. These novel failover constraints introduce new trade-offs not present in classic assignment problems. We design asymptotically optimal online algorithms for this problem in both the worst- and average-case (where demands are drawn i.i.d. from an unknown distribution). Along the way, we will see a lesser-known probabilistic tool: stochastic monotone matchings. Preliminary experiments on real cloud workloads show the potential of our algorithms to generate more power-efficient allocations, which can save millions and significantly reduce the environmental impact of cloud computing.
Algorithm and Incentive Design for Sustainable Resource Allocation: Beyond
Classical Fisher Markets
Technological advances have opened new avenues for designing market
mechanisms for resource allocation, from enhancing resource allocation eDiciency with widespread data availability to enabling real-time algorithm implementation. While these technological advancements hold significant promise, they also introduce new societal challenges pertaining to equity, privacy, data uncertainty, and security that existing market mechanisms often fail to address. My research develops data-driven and online learning algorithms and incentive schemes to address these challenges of traditional market mechanisms, thereby advancing the science and practice of market design for sustainable and society-aware resource allocation.
In this talk, I focus on addressing data uncertainty and privacy issues in the context of Fisher markets, a classical framework for fair resource allocation where the problem of computing equilibrium prices relies on complete information of user attributes, which are typically unavailable in practice. Motivated by this practical limitation, we study a modified online incomplete information variant of Fisher markets, where users with privately known utility and budget parameters, drawn i.i.d. from a distribution, arrive sequentially. In this novel market, we establish the limitations of static pricing and design dynamic posted-price algorithms with improved guarantees. Our main result is a posted-price algorithm that solely
relies on revealed preference (RP) feedback, i.e., observations of user consumption,
achieving the best-known guarantees for first-order algorithms in the RP setting while providing a regret analysis of a fairness-promoting logarithmic objective, unlike typical nonnegative and bounded eDiciency-promoting objectives in online learning.
Link to Paper: https://arxiv.org/pdf/2205.00825
Devansh Jalota is a PhD candidate in Computational and Mathematical Engineering at Stanford University, where he is a Stanford Interdisciplinary Graduate Fellow. His research develops data-driven learning algorithms and incentive schemes to advance the science and practice of market design for sustainable resource allocation, with a particular focus on applications in future mobility systems and electricity markets. Prior to joining Stanford, he received his bachelor’s in applied mathematics and civil engineering at UC Berkeley.
]]>Technological advances have opened new avenues for designing market
mechanisms for resource allocation, from enhancing resource allocation eDiciency with widespread data availability to enabling real-time algorithm implementation. While these technological advancements hold significant promise, they also introduce new societal challenges pertaining to equity, privacy, data uncertainty, and security that existing market mechanisms often fail to address. My research develops data-driven and online learning algorithms and incentive schemes to address these challenges of traditional market mechanisms, thereby advancing the science and practice of market design for sustainable and society-aware resource allocation.
In this talk, I focus on addressing data uncertainty and privacy issues in the context of Fisher markets, a classical framework for fair resource allocation where the problem of computing equilibrium prices relies on complete information of user attributes, which are typically unavailable in practice. Motivated by this practical limitation, we study a modified online incomplete information variant of Fisher markets, where users with privately known utility
and budget parameters, drawn i.i.d. from a distribution, arrive sequentially. In this novel market, we establish the limitations of static pricing and design dynamic posted-price algorithms with improved guarantees. Our main result is a posted-price algorithm that solely relies on revealed preference (RP) feedback, i.e., observations of user consumption, achieving the best-known guarantees for first-order algorithms in the RP setting while providing a regret analysis of a fairness-promoting logarithmic objective, unlike typical nonnegative and bounded eDiciency-promoting objectives in online learning.
Link to Paper: https://arxiv.org/pdf/2205.00825
]]>
Meet MS Urban Analytics program director, Dr. Subhro Guhathakurta, program alumni, and the School of City and Regional Planning academic advisor and graduate program coordinator, Ms. Anna Traykova, to learn more about the program.
Meet MS Urban Analytics program director, Dr. Subhro Guhathakurta, program alumni, and the School of City and Regional Planning academic advisor and graduate program coordinator, Ms. Anna Traykova, to learn more about the program
]]>
Abstract: The use of antiretroviral therapy (ART) has significantly reduced HIV-related mortality and morbidity, transforming HIV infection to a chronic disease with the care now focusing on treatment adherence, comorbidities including mental health, and other long-term outcomes. Since combination ART with three or more drugs of different mechanisms or against different targets is recommended for all people living with HIV (PWH) and they must continue on it indefinitely once started, understanding the long-term ART effects on health outcomes and personalizing ART treatment based on individuals’ characteristics is crucial for optimizing PWH’s health outcomes and facilitating precision medicine in HIV. In this talk, I will present reinforcement learning (RL) methods designed to learn and understand the impact of ART on the health outcomes of PWH, and explore the future of HIV care through incorporating large language models (LLM) with RL.
Bio: Dr. Yanxun Xu is an Associate Professor and Joseph & Suzanne Jenniches Faculty Scholar in the Department of Applied Mathematics and Statistics, Division of Biostatistics in the School of Medicine at Johns Hopkins University. Her research interests lie in developing theory and methods for a broad range of problems, such as reinforcement learning, high-dimensional data analysis, Bayesian nonparametric statistics, and uncertainty quantification. She also develops new statistical and machine learning methods for various applications, including electronic health records, dynamic treatment regimens, cancer genomics, early detection of Alzheimer’s disease, mental health in people with HIV, and early-phase clinical trial designs. Her research has been continuously funded by NSF, NIH, and industries.
]]>
Meet MS Urban Analytics program director, Dr. Subhro Guhathakurta, program alumni, and the School of City and Regional Planning academic advisor and graduate program coordinator, Ms. Anna Traykova, to learn more about the program
]]>Meet MS Urban Analytics program director, Dr. Subhro Guhathakurta, program alumni, and the School of City and Regional Planning academic advisor and graduate program coordinator, Ms. Anna Traykova, to learn more about the program
]]>
Probing-enhanced stochastic programming
Abstract:
We consider a two-stage stochastic program where the decision-maker has the opportunity to obtain information about the distribution of the random variables X through a set of discrete actions that we refer to as probing. Probing allows the decision-maker to observe components of a random vector Y that is jointly-distributed with X. We propose a three-stage optimization model for this problem, where the first-stage variables select components of Y to observe. In the case that X and Y have finite support, a model of Goel and Grossmann can be applied to obtain a formulation of this problem whose size is proportional to the square of cardinality of the sample space of the random variables. We propose to solve the model using bounds obtained from an information-based relaxation, combined with a branching scheme that enforces the consistency of decisions with observed information. The branch-and-bound approach can naturally be combined with sampling in order to estimate both lower and upper bounds on the optimal solution value even for problems with continuous distribution. We demonstrate the approach on instances of a stochastic facility location problem.
This is joint work with Zhichao Ma, Jeff Linderoth, Youngdae Kim, and Logan Matthews.
Bio:
James (Jim) Luedtke is a Professor in the department of Industrial and Systems Engineering at the University of Wisconsin-Madison and a Discovery Fellow at the Wisconsin Institute for Discovery. Luedtke earned his Ph.D. at Georgia Tech and did postdoctoral work at the IBM T.J. Watson Research Center. Luedtke’s research is focused on methods for solving stochastic and mixed-integer optimization problems, as well as applications of such models. His current research interests include investigation of computational methods for solving two and multi-stage stochastic integer programming problems, and integration of optimization and machine learning models. Luedtke serves on the editorial boards of the journals SIAM Journal on Optimization and Mathematical Programming Computation and is chair of the Mathematical Optimization Society Publications Committee.
Georgia Tech ISyE Departmental Seminar Speaker Invitation
]]>
This is joint work with Zhichao Ma, Jeff Linderoth, Youngdae Kim, and Logan Matthews.
Mobilizing Demand Flexibility in Wholesale Electricity Markets with VPP Supply Functions
Abstract:
FERC Order 2222 requires ISOs to develop market mechanisms that will enable aggregators of distributed resources to participate in the electricity wholesale market. In this work we describe the construction of a supply function for an aggregator’s virtual power plant (VPP) based on a portfolio of curtailable devices, categorized into priority tranches, that are controlled by the aggregator through edge technology behind the meter. Only the nameplate capacity of the curtailed devices are known to the aggregator while the energy yield from curtailment is uncertain. However, the supply function of the VPP offered by the aggregator into the wholesale market must specify deliverable energy quantity as function of wholesale price, like any other generator. We employ a revenue management methodology to construct a supply function with controlled delivery risk, based on priority tranches of the curtailed devices and offline estimates of the energy yield probability distributions. This work is part of an ARPA E project aimed at implementing a VPP based on aggregated demand curtailments at PJM.
(Joint work with Hung Po Chao, ETA and Alex Papalexopoulos, ZOME)
Bio:
Dr. Shmuel S. Oren is Professor of the Graduate School in the Department of Industrial Engineering and Operations Research at UC Berkeley and is a co-founder and the Berkeley site director, of PSerc. He has been a member of the California ISO Market Surveillance Committee and a consultant to many private and public entities in the US and abroad. He holds a Ph.D in Engineering Economic Systems from Stanford University. He is a recipient of the INFORMS Hotelling Medal for life time achievement in Energy Natural Resources and Environment and the IEEE Outstanding Power Systems Educator Award. He is a Member of the US National Academy of Engineering, a Life Fellow of the IEEE and Fellow of INFORMS.
]]>Thursday, December 5, 2024 | 12-1pm ET
We will discuss the role of statistical analysis, performance metrics, and advanced tools for supply chain transformation. Practical examples and case studies will be shared to illustrate the impact of these techniques.
]]>Diversity, equity and inclusion and racial and social justice in the field of operations research and analytics: Results from an examination of recent scholarship and university academic programs
Abstract:
This talk describes a project to better understand the state of diversity, equity and inclusion in the decision sciences. By doing so, we can identify areas for improvement in scholarship, practice and education. I discuss two recent studies: a review of recent scholarship in operations research, operations management and supply chain management intended to identify opportunities for DEI-forward work in these fields, and a survey of university programs in OR/analytics to learn about the characteristics of students, faculty, DEI-related curricula and supports for traditionally-underrepresented students.
Co-authors:
Tayo Fabusuyi, Assistant Research Scientist, University of Michigan Transportation Research Institute
Elham Hesari, doctoral candidate, Public Policy PhD program, University of Massachusetts Boston
Bio:
Michael P. Johnson is professor in the Department of Public Policy and Public Affairs and special assistant to the chancellor for Black life at University of Massachusetts Boston. He received his PhD in Operations Research from Northwestern University and his BS from Morehouse College. His research addresses decision modeling for nonprofit organizations and government agencies. His primary application areas include affordable and assisted housing, community development, climate change response, and diversity, equity and inclusion in the decision sciences. He has authored multiple books, including Supporting Shrinkage: Planning and Decision-Making for Legacy Cities (SUNY Press, 2021) and Decision Science for Housing and Community Development: Localized and Evidence‐Based Responses to Distressed Housing and Blighted Communities (Wiley, 2016). He has served as INFORMS vice-president of Chapters and Fora, and was founding chair of the Diversity, Equity and Inclusion committee. His research has received support from the National Science Foundation, the Abell Foundation and the INFORMS Diversity Ambassadors program.
From Academia to Industry: Observations of a Traveler
Marcus Nanotechnology Building, Room 1116-1118
Tuesday, October 29, 2024
3:30-4:30PM
Reception held in the Atrium
Some people meticulously plan every step of their career, mapping out milestones with precision. I’m not one of those people. Instead, I follow where my curiosity leads, seizing opportunities when they present themselves. In transitioning from academia to industry, I've observed how to transform a job into a fulfilling career, the distinct differences between these two worlds, and the importance of continuous learning in the era of AI. I'll share these reflections, and I look forward to hearing your thoughts on them.
Dr. Russell D. Meller is the Chief Scientist at FORTNA, a company that designs and implements complex distribution centers. He oversees a group of 25+ researchers that develop all algorithms in the FortnaWES™ software, perform data analysis on design projects, simulation of designs, emulation of material handling systems and all modeling aspects of the FortnaDCdesign Suite™. Russ holds a B.S., M.S. & Ph.D. in Industrial and Operations Engineering from the University of Michigan. He’s been with FORTNA for over 10 years after an illustrious 20+ year career in academia, holding positions at Auburn University, Virginia Tech, the University of Arkansas and the Technical University of Graz (Austria). In 2020, he was elected to the National Academy of Engineering, the highest honor for an Engineer in the United States. His election was due in large part to creating a scalable design methodology at FORTNA for distribution centers.
Dr. Russell D. Meller is the Chief Scientist at FORTNA, a company that designs and implements complex distribution centers. Russ holds a B.S., M.S. & Ph.D. in Industrial and Operations Engineering from the University of Michigan. He’s been with FORTNA for over 10 years after an illustrious 20+ year career in academia, holding positions at Auburn University, Virginia Tech, the University of Arkansas and the Technical University of Graz (Austria). In 2020, he was elected to the National Academy of Engineering, the highest honor for an Engineer in the United States. His election was due in large part to creating a scalable design methodology at FORTNA for distribution centers.
]]>We strongly encourage you to attend to seek full-time employment, internships, and projects (rather than waiting until the end of the semester).
If you are interested in hosting a table for the upcoming session, please let us know after reviewing the below information within our website. Registration closes Monday, December 30th.
Visit https://www.scl.gatech.edu/outreach/supplychainday.
]]>This course will demonstrate how warehouse operations are a key enabler to a successful supply chain implementation and the starting point for a transformation. It is critical to manage safety, quality and efficiency. Learn to leverage the lean supply chain modifications to improve customer responsiveness and reduce operating costs and in doing so contributing to a supply chain that creates a competitive advantage for a company. To accomplish this goal, we must bring lean principles into the warehouse and distribution center.
The course takes place virtually with via remote classroom instruction October 21 & 25, 8:00am - 5:00pm. Online self-directed work is required October 22, 23, 24.
Supply chain professionals, logistics professionals, material managers, production control managers, transportation managers, warehousing managers and purchasing managers
Upon completion of this course, you will be able to:
Benefits:
In today’s global economy, operating risks are increasingly on the minds of executives. The specific context of operating risk can range from general areas of business continuity to the effects of natural disasters. In this course participants will gain a solid understanding of Supply Chain Risk Management principals including effective ways to identify, mitigate and measure the impact of potential supply chain disruptions.
Upon completion of this course, you will be able to:
Abstract:
This paper investigates the iterates obtained from iterative algorithms in high-dimensional linear regression problems, in the regime where the feature dimension is comparable with the sample size. The analysis and proposed estimators are applicable to Gradient Descent (GD), proximal GD and their accelerated variants such as Fast Iterative Soft-Thresholding (FISTA). The paper proposes novel estimators for the generalization error of the iterate for any fixed iteration along the trajectory. These estimators are proved to be root-n consistent under Gaussian designs. Applications to early-stopping are provided: when the generalization error of the iterates is a U-shape function of the iterations, the estimates allow to select from the data an iteration that achieves the smallest generalization error along the trajectory. Additionally, we provide a technique for developing debiasing corrections and valid confidence intervals for the components of the true coefficient vector from the iterate at any finite iteration.
Bio:
Pierre C Bellec is an associate professor in the Statistics Department at Rutgers University. He was elected fellow of the Institute of Mathematical Statistics in 2023. In the last few years, his work focused on uncertainty quantifications in regression models, including some recent applications to iterative algorithms and bagging. Before joining Rutgers in 2016, he completed his PhD at ENSAE in Paris, France where he was advised by Alexandre Tsybakov.
]]>
This paper investigates the iterates obtained from iterative algorithms in high-dimensional linear regression problems, in the regime where the feature dimension is comparable with the sample size. The analysis and proposed estimators are applicable to Gradient Descent (GD), proximal GD and their accelerated variants such as Fast Iterative Soft-Thresholding (FISTA). The paper proposes novel estimators for the generalization error of the iterate for any fixed iteration along the trajectory. These estimators are proved to be root-n consistent under Gaussian designs. Applications to early-stopping are provided: when the generalization error of the iterates is a U-shape function of the iterations, the estimates allow to select from the data an iteration that achieves the smallest generalization error along the trajectory. Additionally, we provide a technique for developing debiasing corrections and valid confidence intervals for the components of the true coefficient vector from the iterate at any finite iteration.
]]>Convexification and optimization of problems involving the Euclidean norm.
Abstract:
The field of mixed-integer nonlinear optimization has advanced significantly over the past three decades. However, even small instances of many nonconvex optimization problems involving the Euclidean norm are beyond the capabilities of existing algorithms. These problems stem from applications such as molecular energy minimization, object packing, and facility location, and often share specific features that make them challenging. In this presentation, we identify these features, introduce new algorithms to address them, and present numerical results that demonstrate the impact of our techniques. Furthermore, we identify several related open problems and opportunities for analytical and computational advances. This is joint work with Anatoliy Kuznetsov.
Bio:
Nick Sahinidis is Butler Family Chair and Professor of Industrial & Systems Engineering and Chemical & Biomolecular Engineering at the Georgia Institute of Technology. Dr. Sahinidis previously taught at the University of Illinois at Urbana-Champaign (1991-2007) and Carnegie Mellon University (2007-2020). He has pioneered algorithms and developed widely used software for optimization and machine learning. He received the INFORMS Computing Society Prize in 2004, the Beale-Orchard-Hays Prize from the Mathematical Programming Society in 2006, the Computing in Chemical Engineering Award in 2010, the Constantin Carathéodory Prize in 2015, and the National Award and Gold Medal from the Hellenic Operational Research Society in 2016. He is a member of the US National Academy of Engineering, a fellow of INFORMS, a fellow of AIChE, a fellow of the Asia-Pacific Artificial Intelligence Association, and past Editor-in-Chief of Optimization and Engineering.
]]>Abstract:
In this talk, we consider multi-period convex quadratic optimization problems with indicator variables. This problem class has important applications in machine learning and model predictive control. We study a sub-class with a factorable or block-factorable cost matrix and show that it is solvable in polynomial time. We also give a compact convex hull description in an extended space with linear and conic quadratic inequalities. Our computational experiments with data from neuron activation inference and hybrid-electric vehicle power management indicate promises as well as challenges.
Joint work with Andres Gomez and Jisun Lee.
Bio:
Alper Atamturk is the Earl J. Isaac Chair in the Science and Analysis of Decision Making, Professor and Chair of the Department of Industrial Engineering and Operations at the University of California, Berkeley. He received his Ph.D. from the Georgia Institute of Technology in 1998. His research interests are in optimization, integer programming, optimization under uncertainty with applications to machine learning, energy systems, portfolio and network design.
Alper serves as the UC Berkeley site director of the NSF AI Institute for Advances in Optimization. He serves as co-editor for Mathematical Programming, area editor for Mathematical Programming Computation, and associate editor for Operations Research, Discrete Optimization, and Journal of Risk. He is a Fellow of INFORMS and Vannevar Bush Fellow of the US Department of Defense. He received the Farkas Prize from INFORMS Optimization Society in 2023.
Monday, October 7, 2024 | 12-1pm ET
Smarter Decisions for a Better World: Research in Health & INFORMS Innovations
Abstract:
This seminar will contain topics in two major areas. In the first, Dr. Swann will discuss research related to health systems, particularly methodologies from OR, AI, ML and more applied to problems of interest to healthcare providers, governmental agencies, and other decision-makers. Examples include agent-based simulations of pandemics with modeling and assessment of interventions to improve equitable outcomes in the population, learning models to enable dynamic optimization of community policies, and interpretable modeling to identify patients at greatest risk for unplanned hospitalizations.
In the second half of the seminar, Dr. Swann will discuss innovations and future planning with the INFORMS society, including topics such as AI, quantum computing, early career practitioners, accreditation and certification for data science programs, research reproducibility for publications, and strategic planning. Discussion and questions will be welcome.
Bio:
Julie L. Swann is the department head and A. Doug Allison Distinguished Professor of the Edward P. Fitts Department of Industrial and Systems Engineering at NC State University. She is an affiliate faculty in the Joint Department of Biomedical Engineering at NC State and the University of North Carolina at Chapel Hill. Previously, Swann was the Nash Professor in the Stewart School of Industrial and Systems at Georgia Tech, a co-founder of the Center of Health and Humanitarian Systems at Georgia Tech, and the Chair of the Council of Industrial Engineering Academic Department Heads (CIEADH). Swann is a Fellow of INFORMS, IISE, and AIMBE and the 2024 President of INFORMS.
Throughout her career, Swann has conducted research, outreach and education to improve how health and humanitarian systems and supply chains operate worldwide while advancing scientific innovations. Her research relates to public health, public policy, epidemiology, infectious disease, supply chain management, and disaster response, along with building or employing mathematical and computational models. The work allowed her to serve as a science advisor for the H1N1 pandemic response at the Centers for Disease Control and Prevention. From 2020-2023 she was active in supporting pandemic preparations and decision-making in local, state, and federal health agencies, while also serving as a subject matter expert to the media including outlets such as The Washington Post, The LA Times, The Atlantic, The Wall Street Journal, Forbes, Scientific American, and The Hill) along with radio and television.
Along with the CDC, Swann has collaborated with health and humanitarian organizations such as The American Red Cross; The Carter Center; CARE USA; Children’s Healthcare of Atlanta; Emory University Hospital; State Departments of Public Health; and many others including companies.
Worldwide, Dr. Swann has contributed to the education of thousands of practitioners in health and humanitarian systems through the co-creation and teaching in a professional certificate program at Georgia Tech. This contribution includes teaching in the MASHLM program in Lugano, Switzerland, and co-chairing the annual Health and Humanitarian Logistics Conference.
]]>This seminar will contain topics in two major areas. In the first, Dr. Swann will discuss research related to health systems, particularly methodologies from OR, AI, ML and more applied to problems of interest to healthcare providers, governmental agencies, and other decision-makers. Examples include agent-based simulations of pandemics with modeling and assessment of interventions to improve equitable outcomes in the population, learning models to enable dynamic optimization of community policies, and interpretable modeling to identify patients at greatest risk for unplanned hospitalizations.
In the second half of the seminar, Dr. Swann will discuss innovations and future planning with the INFORMS society, including topics such as AI, quantum computing, early career practitioners, accreditation and certification for data science programs, research reproducibility for publications, and strategic planning. Discussion and questions will be welcome.
]]>Multi-agent higher-order learning vs Nash equilibrium
Abstract:
The framework of multi-agent learning explores the dynamics of how individual agent strategies evolve in response to the evolving strategies of other agents. Of particular interest is whether agent strategies converge to well-known solution concepts such as Nash Equilibrium (NE). Most “fixed order” learning dynamics restrict an agent’s underlying state to be its own strategy. In “higher order” learning, agent dynamics can include auxiliary states that can capture phenomena such as path dependencies. We introduce higher-order gradient play dynamics that resemble projected gradient ascent with auxiliary states. The dynamics are “payoff based” in that each agent's dynamics depend on its own evolving payoff. While these payoffs depend on the strategies of other agents in a game setting, agent dynamics do not depend explicitly on the nature of the game or the strategies of other agents. In this sense, dynamics are “uncoupled” since an agent’s dynamics do not depend explicitly on the utility functions of other agents. We first show that for any specific game with an isolated completely mixed-strategy NE, there exist higher-order gradient play dynamics that lead (locally) to that NE, both for the specific game and nearby games with perturbed utility functions. Conversely, we show that for any higher-order gradient play dynamics, there exists a game with a unique isolated completely mixed-strategy NE for which the dynamics do not lead to NE. These results build on prior work that showed that uncoupled fixed-order learning cannot lead to NE in certain instances, whereas higher-order variants can. Finally, we consider the mixed-strategy equilibrium associated with coordination games. While higher-order gradient play can converge to such equilibria, we show such dynamics must be inherently irrational.
Bio:
Jeff S. Shamma is with the University of Illinois at Urbana-Champaign where he is the Department Head of Industrial and Enterprise Systems Engineering (ISE) and Jerry S. Dobrovolny Chair. His prior academic appointments include faculty positions at the King Abdullah University of Science and Technology (KAUST) and the Georgia Institute of Technology, where he was the Julian T. Hightower Chair in Systems and Controls. Jeff received a PhD in Systems Science and Engineering from MIT in 1988. He is a Fellow of IEEE and IFAC; a recipient of the IFAC High Impact Paper Award, AACC Donald P. Eckman Award, and NSF Young Investigator Award; and a past Distinguished Lecturer of the IEEE Control Systems Society. He has been a plenary or semi-plenary speaker at several conferences, including NeurIPS, World Congress of the Game Theory Society, IEEE Conference on Decision and Control, and the American Control Conference. Jeff is currently serving as the Editor-in-Chief for the IEEE Transactions on Control of Network Systems
]]>Friday, September 13, 2024 | 12-1pm ET
Thursday, September 19, 2024 | 12-1pm ET
Thursday, August 29, 2024 | 12-1pm ET
Thursday, September 3, 2024 | 12-1pm ET
This course is the fourth in the 4-course Supply Chain Analytics Professional certificate program. It incorporates learning advanced analytics and mathematical optimization to find solutions for supply chain problems. You’ll learn how to use linear programming, mixed integer programming, and heuristics to conduct prescriptive analytics related to production processes, distribution networks, and routing. The course serves as a capstone for the program by culminating in a hackathon where you’ll design networks, inventory policies, and scenarios and then evaluate the outcomes via simulations.
The online version of the course is comprised of (4) half-day online instructor-led LIVE group webinars and pre-work (e.g. installing and testing software on your computer, testing connectivity with LMS and meeting software, etc.) to be completed before the first day of the course.
Experienced business professionals who perform or want to perform analytics to improve their supply chain management processes. They want to tackle strategic goals and to perform leading edge analytics projects that address the full complexity of supply chains.
This course prepares students in the basics of transportation operations and analysis. The course includes review of the key elements of transportation such as: modes of transportation, transportation procurement, cost minimization techniques, the role of ports in global logistics, and international trade terms. The course also will discuss emerging trends in North American transportation markets, emerging techniques, and greenhouse gas emissions reduction.
This course is designed for Supply Chain Managers, Distribution Managers, Transportation Planners, Transportation Clerks, Transportation Analysts, and Transportation Managers and learners seeking to enter these roles. Supply chain professionals from other domains will also benefit through gaining insights into transportation operations.
Upon completion of this course, you will be able to:
We strongly encourage you to attend to seek full-time employment, internships, and projects (rather than waiting until the end of the semester).
If you are interested in hosting a table for the upcoming session, please let us know after reviewing the below information within our website. Registration closes Monday September 2nd.
Visit https://www.scl.gatech.edu/outreach/supplychainday.
]]>Trustworthy Forecasting Algorithms
ABSTRACT:
Algorithms are increasingly tasked with forecasting the probabilities of uncertain events: a creditor repaying a loan, a user clicking an advertisement, or a word appearing next in a stream of text, for example. Such forecasts are trustworthy if their users can be sure they won't regret treating the predicted probabilities as if they were the actual distributions from which outcomes were sampled. The term "calibration" refers to various measures of forecast accuracy that attempt to formalize this property of trustworthiness. Defining calibration, and designing algorithms to achieve it, turns out to be a tightrope walk between strong definitions, which ensure reliable results for downstream users but are computationally and statistically harder to achieve, and weak definitions, which have the opposite benefits and drawbacks. I will report on some recent research that locating a sweet spot between these two extremes, requiring no more samples or computation than the weakest definitions but providing guarantees that are, in many cases, as useful for downstream users as the strongest ones.
This talk is based on joint work with Michael Kim, Princewill Okoroafor, Renato Paes Leme, Jon Schneider, and Yifeng Teng.
BIO:
Bobby Kleinberg is a Professor of Computer Science at Cornell University and a part-time Faculty Researcher at Google. His research concerns algorithms and their applications to machine learning, economics, networking, and other areas. Prior to receiving his doctorate from MIT in 2005, Kleinberg spent three years at Akamai Technologies; he and his co-workers received the 2018 SIGCOMM Networking Systems Award for pioneering the first Internet content delivery network. He is a Fellow of the ACM and a recipient of the ACM SIGecom Mid-Career Award for advancing the understanding of on-line learning and decision problems and their application to mechanism design.