{"689162":{"#nid":"689162","#data":{"type":"event","title":"PhD Defense by Sourabh Choudhary","body":[{"value":"\u003Cp\u003ETitle: Tractable Reformulations and Heuristic Methods for Challenging Mixed-Integer Nonlinear Programs\u003Cbr\u003E\u003Cbr\u003EDate: April 3rd, 2026\u003Cbr\u003ETime: 10:00 am \u2013 12:00 pm (ET)\u003Cbr\u003ELocation: Groseclose 118 and Teams: https:\/\/teams.microsoft.com\/meet\/28137700678182?p=0C21ygpjkwlKbhv1xR\u003C\/p\u003E\u003Cp\u003E\u003Cbr\u003ESourabh K. Choudhary\u003Cbr\u003EPhD Candidate in OR\u003Cbr\u003EISyE, Georgia Tech\u003Cbr\u003E\u003Cbr\u003EThesis Committee\u003Cbr\u003E1 Dr. Nick Sahinidis (ISYE, Georgia Tech) (Advisor)\u003Cbr\u003E2 Dr. Santanu Dey (ISYE, Georgia Tech) (Advisor)\u003Cbr\u003E3 Dr. Oktay Gunluk (ISYE, Georgia Tech)\u003Cbr\u003E4 Dr. Alejandro Toriello (ISYE, Georgia Tech)\u003Cbr\u003E5 Dr. Joseph Scott (ChBE, Georgia Tech)\u003Cbr\u003E\u003Cbr\u003EAbstract\u003Cbr\u003EMany real-world decision problems can be formulated as mixed-integer nonlinear programs (MINLPs), but solving such problems to high quality remains challenging due to nonconvexities, combinatorial structure, and large-scale formulations. This dissertation investigates how linear programming (LP) and mixed-integer linear programming (MILP) techniques can be leveraged to address difficult classes of nonconvex optimization problems.\u003C\/p\u003E\u003Cp\u003EFirst, I examine binary polynomial optimization (BPO) through the lens of structural sparsity. In particular, I develop LP formulations based on treewidth of the underlying graph structure, showing how exploiting problem structure can lead to smaller LP equivalents.\u003C\/p\u003E\u003Cp\u003ENext, I study water network optimization problems, which are important in the design and operation of infrastructure systems and are often modeled as difficult nonconvex MINLPs. I show how mixed integer linear reformulations and MILP-guided strategies can help produce strong feasible solutions and improve computational tractability on these practically relevant instances.\u003C\/p\u003E\u003Cp\u003EFinally, I present a primal heuristic for MINLP, called Contraction Aided Search (CAS), which uses solutions of mixed-integer linear relaxations\/approximations to guide the search for high-quality feasible solutions of the original mixed integer nonlinear problem. The approach is particularly effective for challenging mixed-integer quadratically constrained quadratic programs (MIQCQPs), where finding strong feasible solutions early can improve solvers\u0027 practical performance.\u003C\/p\u003E\u003Cp\u003EOverall, this dissertation highlights the power of LP- and MILP-based methods as a unifying framework for tackling challenging nonconvex optimization problems.\u003Cbr\u003E\u0026nbsp;\u003C\/p\u003E","summary":"","format":"limited_html"}],"field_subtitle":"","field_summary":[{"value":"\u003Cp\u003ETractable Reformulations and Heuristic Methods for Challenging Mixed-Integer Nonlinear Programs\u003C\/p\u003E","format":"limited_html"}],"field_summary_sentence":[{"value":"Tractable Reformulations and Heuristic Methods for Challenging Mixed-Integer Nonlinear Programs"}],"uid":"27707","created_gmt":"2026-03-24 18:13:22","changed_gmt":"2026-03-24 18:13:59","author":"Tatianna Richardson","boilerplate_text":"","field_publication":"","field_article_url":"","field_event_time":{"event_time_start":"2026-04-03T10:00:00-04:00","event_time_end":"2026-04-03T12:00:00-04:00","event_time_end_last":"2026-04-03T12:00:00-04:00","gmt_time_start":"2026-04-03 14:00:00","gmt_time_end":"2026-04-03 16:00:00","gmt_time_end_last":"2026-04-03 16:00:00","rrule":null,"timezone":"America\/New_York"},"location":"Groseclose 118 and Teams","extras":[],"groups":[{"id":"221981","name":"Graduate Studies"}],"categories":[],"keywords":[{"id":"185896","name":"PhD Defen"}],"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":""}}}