{"681803":{"#nid":"681803","#data":{"type":"event","title":"PhD Proposal by Kartik Sharma","body":[{"value":"\u003Cp\u003ETitle: Towards reliability and controllability of machine learning models\u003C\/p\u003E\u003Cp\u003EDate: Monday, April 28, 2025\u003Cbr\u003ETime: 1:00pm - 3:00pm ET\u003Cbr\u003ELocation: (hybrid) CODA C1308 Cabbagetown and Zoom https:\/\/gatech.zoom.us\/j\/98399801627?pwd=rN37t8hs0ErO38Us0xIOVywtzUbKaE.1\u003C\/p\u003E\u003Cp\u003EKartik Sharma\u003Cbr\u003ESchool of Computational Science and Engineering\u003Cbr\u003ECollege of Computing\u003Cbr\u003EGeorgia Institute of Technology\u003C\/p\u003E\u003Cp\u003EWebsite: https:\/\/ksartik.github.io\/\u003C\/p\u003E\u003Cp\u003E\u003Cbr\u003ECommittee:\u003Cbr\u003EDr. Srijan Kumar, Advisor (School of Computational Science and Engineering, Georgia Institute of Technology)\u003Cbr\u003EDr. Bo Dai (School of Computational Science and Engineering, Georgia Institute of Technology)\u003Cbr\u003EDr. Chao Zhang (School of Computational Science and Engineering, Georgia Institute of Technology)\u0026nbsp;\u003Cbr\u003EDr. Rakshit Trivedi (Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology)\u003C\/p\u003E\u003Cp\u003EAbstract:\u003Cbr\u003EDespite their remarkable capabilities, the widespread adoption of machine learning (ML) systems is hindered by their unpredictability and lack of user control. This dissertation addresses these challenges by advancing the reliability and controllability of ML models across three key directions:\u003Cbr\u003E1.\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;User-controllable generation: A projected sampling method, PRODIGY, is introduced for generating graphs\/drugs from pre-trained graph diffusion models under hard structural constraints (e.g., molecular weight, valency) without retraining. To support domain-expert control in retrieval-augmented generation, a hypergraph-based retrieval method, OG-RAG is designed to ground query-relevant contexts in domain-specific ontologies.\u003Cbr\u003E2.\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;Measuring reliability: Practical reliability gaps are exposed in dynamic graph neural networks (GNNs) by formulating undetectable perturbations using a novel constraint and demonstrating novel vulnerabilities with an efficient PGD-based approach. A systematic framework, ELK, is also proposed to efficiently identify factuality gaps in large language models (LLMs) by estimating their knowledge using sentence embeddings.\u003Cbr\u003E3.\u0026nbsp;\u0026nbsp;\u0026nbsp;\u0026nbsp;Robust and domain-specific architectures: By leveraging sociological theories of structural balance, SEMBA is designed to learn effective representations of nodes in dynamic signed social networks. Another contribution, MetSelect, enhances the current architectures of GNNs by learning to select a node-personalized representation layer to classify each node.\u0026nbsp;\u003Cbr\u003EFuture work extends these directions to language and multimodal data, focusing on LLM robustness via adaptive system prompt mechanism, imitation learning agents with inner speech to better capture human behavioral diversity, and reinforcement learning-based token pruning for memory-constrained LLM reasoning.\u003Cbr\u003E\u0026nbsp;\u003C\/p\u003E","summary":"","format":"limited_html"}],"field_subtitle":"","field_summary":[{"value":"\u003Cp\u003ETowards reliability and controllability of machine learning models\u003C\/p\u003E","format":"limited_html"}],"field_summary_sentence":[{"value":"Towards reliability and controllability of machine learning models"}],"uid":"27707","created_gmt":"2025-04-15 17:58:02","changed_gmt":"2025-04-15 17:58:39","author":"Tatianna Richardson","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2025-04-28T13:00:00-04:00","event_time_end":"2025-04-28T15:00:00-04:00","event_time_end_last":"2025-04-28T15:00:00-04:00","gmt_time_start":"2025-04-28 17:00:00","gmt_time_end":"2025-04-28 19:00:00","gmt_time_end_last":"2025-04-28 19:00:00","rrule":null,"timezone":"America\/New_York"},"location":"(hybrid) CODA C1308 Cabbagetown and Zoom ","extras":[],"groups":[{"id":"221981","name":"Graduate Studies"}],"categories":[],"keywords":[{"id":"102851","name":"Phd proposal"}],"core_research_areas":[],"news_room_topics":[],"event_categories":[{"id":"1788","name":"Other\/Miscellaneous"}],"invited_audience":[{"id":"78771","name":"Public"}],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[],"email":[],"slides":[],"orientation":[],"userdata":""}}}