{"id":12076,"date":"2026-07-20T10:17:13","date_gmt":"2026-07-20T15:17:13","guid":{"rendered":"https:\/\/meetings.informs.org\/wordpress\/annual\/?page_id=12076"},"modified":"2026-08-17T06:38:58","modified_gmt":"2026-08-17T11:38:58","slug":"tutorials","status":"publish","type":"page","link":"https:\/\/meetings.informs.org\/wordpress\/annual\/tutorials\/","title":{"rendered":"TutORials"},"content":{"rendered":"<!--themify_builder_content-->\n<div id=\"themify_builder_content-12076\" data-postid=\"12076\" class=\"themify_builder_content themify_builder_content-12076 themify_builder tf_clear\">\n                    <div  data-zoom-bg=\"desktop\" data-css_id=\"mpfz468\" data-lazy=\"1\" class=\"module_row themify_builder_row fullwidth_row_container tb_mpfz468 tb_first tf_w\">\n            <span  class=\"builder_row_cover tf_abs\" data-lazy=\"1\"><\/span>            <div class=\"row_inner col_align_top tb_col_count_1 tf_box tf_rel\">\n                        <div  data-lazy=\"1\" class=\"module_column tb-column col-full tb_4m4r468 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_rmqh468   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <h1 class=\"page-title\">TutORials<\/h1>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                        <\/div>\n        <\/div>\n                        <div  data-lazy=\"1\" class=\"module_row themify_builder_row tb_tn6t430 tf_w\">\n                        <div class=\"row_inner col_align_top tb_col_count_1 tf_box tf_rel\">\n                        <div  data-lazy=\"1\" class=\"module_column tb-column col-full tb_isy5430 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_xxfv430   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p>The <em>TutORials in Operations Research<\/em> series is published annually by INFORMS as an introduction to emerging and classical subfields of operations research and management science. These chapters are designed to be accessible for all constituents of the INFORMS community, including current students, practitioners, faculty, and researchers. The publication allows readers to keep pace with new developments in the field and serves as augmenting material for a selection of the tutorial presentations offered at the INFORMS Annual Meeting.<\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                        <\/div>\n        <\/div>\n                        <div  data-lazy=\"1\" class=\"module_row themify_builder_row tb_lfzn531 tf_w\">\n                        <div class=\"row_inner col_align_top tb_col_count_2 tf_box tf_rel\">\n                        <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_zt5r531 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_gk10531   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p><strong>Sunday, 8:00 AM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_bnk9531 last\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_zrww325   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <h3>Transform Method for Stochastic Processing and Matching Networks<\/h3>\n<p><strong>Speakers: Sushil\u202fVarma, Prakirt Jhunjhunwala, Daniela Hurtado-Lange, Siva <br>Theja Maguluri<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text --><!-- module accordion -->\n<div  class=\"module module-accordion tb_2dp979 \" data-behavior=\"toggle\" data-lazy=\"1\">\n    \n    <ul class=\"ui module-accordion   tb_default_color\">\n            <li>\n            <div class=\"accordion-title tf_rel\">\n                <a href=\"#acc-2dp979-0\" class=\"tb_title_accordion\" aria-controls=\"acc-2dp979-0-content\" aria-expanded=\"false\">\n                    <i class=\"accordion-icon\"><svg  class=\"tf_fa tf-ti-plus\" aria-hidden=\"true\"><use href=\"#tf-ti-plus\"><\/use><\/svg><\/i>                    <i class=\"accordion-active-icon tf_hide\"><svg  class=\"tf_fa tf-ti-minus\" aria-hidden=\"true\"><use href=\"#tf-ti-minus\"><\/use><\/svg><\/i>                    <span class=\"accordion-title-wrap\">Presentation and Speaker Bios<\/span>                <\/a>\n            <\/div><!-- .accordion-title -->\n            <div id=\"acc-2dp979-0-content\" data-id=\"acc-2dp979-0\" aria-hidden=\"true\" class=\"accordion-content tf_hide tf_clearfix\">\n                        <div  data-lazy=\"1\" class=\"module_subrow themify_builder_sub_row tf_w col_align_top tb_col_count_1 tb_lm5v106\">\n                <div  data-lazy=\"1\" class=\"module_column sub_column tb_kc8p106 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_a01s106   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <div class=\"ewa-rteLine\">Modern service systems\u2014ranging from cloud data centers and ride-hailing platforms to healthcare facilities\u2014operate at massive scales where congestion is a critical challenge. Utilizing an operations research approach, these systems are analyzed by modeling them as complex stochastic processes, which are typically understood through process-level convergence to fluid and diffusion limits. However, these methods often prove technically dense and provide limited guidance for finite, practical system scales.<\/div>\n<div>\u00a0<\/div>\n<div class=\"ewa-rteLine\">The transform method, presented in this tutorial, was recently developed as a unified and tractable framework for the steady-state analysis of Stochastic Processing and Matching Networks (SPNs\/SMNs). The transform method overcomes the technical hurdles\u00a0of process-level convergence by working directly with the pre-limit system. By exploiting the zero-drift property of exponential test functions, the method derives explicit functional equations (acting as a proxy for global balance equations) for the transforms (such as moment-generating functions) of queue-length distributions. This approach provides sharp, non-asymptotic performance guarantees, bridging the gap between theoretical asymptotics and real-world system behavior. Since its introduction in 2020 for load-balancing in data center networks, the transform method has been extended to handle realistic complexities, including customer abandonment, state-dependent arrivals, Markov-modulated arrivals, large-system scale, and multi-dimensional networks with multiple bottlenecks. We survey these theoretical advances and demonstrate their practical relevance across diverse domains, such as matching markets and networked service systems. The transform method provides interpretable bounds tied directly to system parameters, offering a powerful analytical alternative to simulation-heavy or purely asymptotic approaches for system design and control.<\/div>\n<div>\u00a0<\/div>\n<div>\n<div class=\"ewa-rteLine\"><strong>Sushil Mahavir Varma<\/strong> is an assistant professor in the Industrial and Operations Engineering Department at the University of Michigan, Ann Arbor. Before joining Michigan, he was a postdoctoral researcher at INRIA Paris. He received his PhD degree in operations research from Georgia Institute of Technology. His research lies broadly in applied probability. He has worked on problems spanning two-sided matching markets, electric vehicle operations, queueing theory, load balancing, and graph alignment. His work has been recognized with the 2024 ACM SIGMETRICS Doctoral Dissertation Award, the 2025 Georgia Tech Sigma Xi <br>Best PhD Thesis Award, and finalist recognition for the 2025 INFORMS TSL Dissertation Award.<\/div>\n<div class=\"ewa-rteLine\">\u00a0<\/div>\n<div class=\"ewa-rteLine\"><strong>Siva Theja Magulur<\/strong>i is Fouts Family Early Career Professor and Associate Professor in the H. Milton Stewart School of Industrial and Systems Engineering at Georgia Tech. He received his B.Tech in Electrical Engineering from IIT Madras, M.S in ECE, M.S. in Applied Math and a PhD in ECE all from University of Illinois at Urbana Champaign. His research interests span the areas of Networks, Control, Optimization, Algorithms, Applied Probability and Reinforcement Learning. He is a recipient of the biennial \u201cBest Publication in Applied Probability\u201d award, NSF CAREER award, \u201cCTL\/BP Junior Faculty Teaching Excellence Award,\u201d and \u201cStudent Recognition of Excellence in Teaching: Class of 1934 CIOS Award.\u201d<\/div>\n<div class=\"ewa-rteLine\">\u00a0<\/div>\n<div class=\"ewa-rteLine\"><strong>Prakirt Jhunjhunwala<\/strong> is a Postdoctoral Scientist at Amazon in the FBA Science team. Previously, he was a Postdoctoral Research Scholar at Columbia Business School, New York. He received his Ph.D. in Operations Research and a Masters in Mathematics from Georgia Tech in 2023 and his B.Tech. (Honors) in Electrical Engineering from IIT Bombay. His research focuses on the design and analysis of stochastic networks, with applications in data centers and quantum networks. Prakirt received an Honorable Mention for the SIGMETRICS Doctoral Dissertation Award (2023), the Best Paper Award at SPCOM (2018), and the Ed Iacobucci Fellowship for Excellence in Applied Probability and Simulation (2022).<\/div>\n<div class=\"ewa-rteLine\">\u00a0<\/div>\n<div class=\"ewa-rteLine\"><strong>Daniela Hurtado-Lange<\/strong> is an assistant professor in the Operations Department at the Kellogg School of Management at Northwestern University. Before joining Kellogg, she was an assistant professor of Mathematics at William &amp; Mary for 1.5 years. She obtained her Ph.D. in Operations Research from Georgia Tech in December 2021, and her research focuses on performance analysis of stochastic processing networks. Specifically, she works on heavy-traffic analysis and queueing theory. Her work has been recognized with the 2022 Sigma Xi Best Ph.D. Thesis Award, and second place in the 2020 JFIG competition.<\/div>\n<\/div>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <\/div>\n                    <\/div><!-- .accordion-content -->\n        <\/li>\n        <\/ul>\n\n<\/div><!-- \/module accordion -->        <\/div>\n                        <\/div>\n        <\/div>\n                        <div  data-lazy=\"1\" class=\"module_row themify_builder_row tb_trb6383 tf_w\">\n                        <div class=\"row_inner col_align_top tb_col_count_2 tf_box tf_rel\">\n                        <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_m1sz383 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_syl8383   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p><strong>Sunday, 11:00 AM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_f2sp383 last\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_hh2d739   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <h3>Deep Learning for Sequential Decision Making under Uncertainty: Foundations, Frameworks, and Frontiers<\/h3>\n<p><strong>Speaker: Esra B\u00fcy\u00fcktahtak\u0131n Toy<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text --><!-- module accordion -->\n<div  class=\"module module-accordion tb_1g8f204 \" data-behavior=\"toggle\" data-lazy=\"1\">\n    \n    <ul class=\"ui module-accordion   tb_default_color\">\n            <li>\n            <div class=\"accordion-title tf_rel\">\n                <a href=\"#acc-1g8f204-0\" class=\"tb_title_accordion\" aria-controls=\"acc-1g8f204-0-content\" aria-expanded=\"false\">\n                    <i class=\"accordion-icon\"><svg  class=\"tf_fa tf-ti-plus\" aria-hidden=\"true\"><use href=\"#tf-ti-plus\"><\/use><\/svg><\/i>                    <i class=\"accordion-active-icon tf_hide\"><svg  class=\"tf_fa tf-ti-minus\" aria-hidden=\"true\"><use href=\"#tf-ti-minus\"><\/use><\/svg><\/i>                    <span class=\"accordion-title-wrap\">Presentation and Speaker Bio<\/span>                <\/a>\n            <\/div><!-- .accordion-title -->\n            <div id=\"acc-1g8f204-0-content\" data-id=\"acc-1g8f204-0\" aria-hidden=\"true\" class=\"accordion-content tf_hide tf_clearfix\">\n                        <div  data-lazy=\"1\" class=\"module_subrow themify_builder_sub_row tf_w col_align_top tb_col_count_1 tb_llvx224\">\n                <div  data-lazy=\"1\" class=\"module_column sub_column tb_h3v9224 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_08dp224   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p>Artificial intelligence (AI) is moving beyond prediction toward systems supporting decisions in complex, dynamic environments. This shift creates a natural<br>intersection with operations research and management science (OR\/MS), which<br>has long provided methodological foundations for sequential decision making under uncertainty. At the same time, deep learning advances, including feedforward neural networks, recurrent architectures, transformers, large language models (LLMs), and deep reinforcement learning, have expanded data-driven modeling for large-scale decisions.<\/p>\n<p>This tutorial presents an OR\/MS-centered perspective on deep learning for sequential decision making under uncertainty, bridging neural architectures and OR\/MS approaches to decision making. Its premise: deep learning complements optimization rather than replacing it. Deep learning brings adaptability and scalable approximation, whereas OR\/MS provides the mathematical rigor to represent constraints, recourse, uncertainty, and decision quality. The tutorial reviews key decision making foundations, connects them to the major neural architectures in modern AI, and organizes the field around three central themes: predict-then-optimize and decision-aware\u00a0learning, learning-based decision generation under constraints for continuous\u00a0and discrete problems with temporal coupling, and deep reinforcement learning for\u00a0sequential and combinatorial decision making. Impact spans supply chains, service\u00a0systems, healthcare and epidemic response, agriculture, energy, environmental sustainability,\u00a0and autonomous operations. This tutorial frames these developments as\u00a0part of a shift from predictive AI toward decision-capable AI, highlighting OR\/MS\u2019s\u00a0role in shaping the next generation of integrated learning\u2013optimization systems.<\/p>\n<p><strong>Esra B\u00fcy\u00fcktahtak\u0131n<\/strong>\u00a0is a Full Professor of Operations Research in the Grado Department of Industrial and Systems Engineering at Virginia Tech, where she directs the Systems\u00a0Optimization and Machine Learning Lab (SysOptiMaL). She earned her Ph.D. in Operations\u00a0Research from the University of Florida. Her research advances multi-stage stochastic mixed-integer\u00a0programming through optimization theory, algorithmic innovation, and learning-enhanced\u00a0optimization. She is recognized for pioneering contributions at the interface of deep learning and\u00a0optimization for sequential decision making under uncertainty. Her work bridges operations research,\u00a0machine learning, and artificial intelligence, with applications in health systems, epidemic\u00a0supply chains, biosecurity, defense, ecological conservation, agriculture, forestry, and environmental\u00a0sustainability.<\/p>\n<p>Dr. B\u00fcy\u00fcktahtak\u0131n is the recipient of the National Science Foundation (NSF) CAREER Award (2016) and the INFORMS Minority Issues Forum (MIF) Early Career Award (2016), and her research\u00a0has been supported by NSF, the U.S. Department of Agriculture (USDA), the Office of Naval Research (ONR), the U.S. Forest Service, the Virginia Department of Forestry, the Minnesota\u00a0Aquatic Invasive Species Research Center (MAISRC), and the 4-VA Collaborative Research\u00a0Program, with more than $3 million in external funding. She has authored 49 peer-reviewed journal\u00a0publications in leading scholarly outlets, including 27 papers in A* and A ranked journals. Her\u00a0work has received six INFORMS Best Publication Awards and has been featured three times in\u00a0ISE Magazine, as well as in the INFORMS Computing Society and U.S. Forest Service newsletters.\u00a0Her professional service includes leadership in the INFORMS community, where she served\u00a0as President of the INFORMS Junior Faculty Interest Group (JFIG, 2014\u20132015) and received two\u00a0INFORMS Service <br>Awards. She currently serves as an Associate Editor of the INFORMS Journal\u00a0<br>on Computing.<\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <\/div>\n                    <\/div><!-- .accordion-content -->\n        <\/li>\n        <\/ul>\n\n<\/div><!-- \/module accordion -->        <\/div>\n                        <\/div>\n        <\/div>\n                        <div  data-lazy=\"1\" class=\"module_row themify_builder_row tb_2sp3149 tf_w\">\n                        <div class=\"row_inner col_align_top tb_col_count_2 tf_box tf_rel\">\n                        <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_yv7s149 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_6r16149   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p><strong>Sunday, 1:15 PM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_3xq7149 last\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_n519272   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <h3>A Tutorial on Reinforcement Learning for LLMs: RLHF and Beyond<\/h3>\n<p><strong>Speaker: Daniel Jiang<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text --><!-- module accordion -->\n<div  class=\"module module-accordion tb_64os577 \" data-behavior=\"toggle\" data-lazy=\"1\">\n    \n    <ul class=\"ui module-accordion   tb_default_color\">\n            <li>\n            <div class=\"accordion-title tf_rel\">\n                <a href=\"#acc-64os577-0\" class=\"tb_title_accordion\" aria-controls=\"acc-64os577-0-content\" aria-expanded=\"false\">\n                    <i class=\"accordion-icon\"><svg  class=\"tf_fa tf-ti-plus\" aria-hidden=\"true\"><use href=\"#tf-ti-plus\"><\/use><\/svg><\/i>                    <i class=\"accordion-active-icon tf_hide\"><svg  class=\"tf_fa tf-ti-minus\" aria-hidden=\"true\"><use href=\"#tf-ti-minus\"><\/use><\/svg><\/i>                    <span class=\"accordion-title-wrap\">Presentation and Speaker Bio<\/span>                <\/a>\n            <\/div><!-- .accordion-title -->\n            <div id=\"acc-64os577-0-content\" data-id=\"acc-64os577-0\" aria-hidden=\"true\" class=\"accordion-content tf_hide tf_clearfix\">\n                        <div  data-lazy=\"1\" class=\"module_subrow themify_builder_sub_row tf_w col_align_top tb_col_count_1 tb_1gy3597\">\n                <div  data-lazy=\"1\" class=\"module_column sub_column tb_wjy6597 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_ez3j597   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p>Large language models (LLMs) require post-training, which takes a foundation model and trains it to follow instructions, behave safely, and perform well across downstream use cases. Reinforcement learning (RL) post-training has become one of the\u00a0most common post-training approaches, most prominently reinforcement learning from human feedback (RLHF), where the reward reflects human preference judgments. In this tutorial, we formulate RL post-training as a sequential decision problem (a Markov decision process), covering both single-turn settings, where the model produces one response to a prompt, and multi-turn settings,\u00a0where it interacts with a user or environment over multiple rounds. We also distinguish reward signals by how directly they can be measured or inferred (verifiable, observable, or latent). We then discuss reward overoptimization, a phenomenon in which the policy exploits errors in a learned reward model, motivating a KL-regularized objective that keeps the policy close to a reference model while still optimizing reward. From this objective, we give precise step-by-step derivations of four central policy-optimization algorithms (DPO, REINFORCE, PPO, and GRPO), explicitly distinguishing exact derivations from practical approximations and heuristics.<\/p>\n<p>The tutorial also covers deployment paradigms, with a focus on batch-online deployment, in which a policy\u2019s interaction\u00a0data is periodically collected and used for offline retraining and subsequent redeployment. For multi-turn settings, we cover an\u00a0extension of GRPO to full trajectories collected from a training environment, and a batch-online method that performs approximate\u00a0policy iteration from logged trajectories. Finally, we discuss open research directions in RL post-training that may benefit from a\u00a0broad range of research perspectives.<\/p>\n<p><strong>Daniel R. Jiang<\/strong> is a Research Scientist at Meta and an Adjunct Professor of Industrial Engineering at the University of Pittsburgh. His research spans reinforcement learning, sequential decision-making, and adaptive experimentation, with a recent emphasis on reinforcement learning for LLM post-training. His work contributed to the first real-world deployment of an RL-trained language model for generative advertising on Facebook and to the training of conversational agents through multi-turn RLHF. He has also developed multi-step lookahead methods for adaptive experimentation and co-created BoTorch, a widely used open-source library for Bayesian optimization in PyTorch. His publications have appeared in journals including Management Science, Operations Research, JMLR, and Mathematics of Operations Research, and at conferences such as NeurIPS, ICML, and ICLR. Daniel received a Ph.D. in Operations Research and Financial Engineering from Princeton University and dual B.S. degrees in Electrical and Computer Engineering and Mathematics from Purdue University.<\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <\/div>\n                    <\/div><!-- .accordion-content -->\n        <\/li>\n        <\/ul>\n\n<\/div><!-- \/module accordion -->        <\/div>\n                        <\/div>\n        <\/div>\n                        <div  data-lazy=\"1\" class=\"module_row themify_builder_row tb_g6ej384 tf_w\">\n                        <div class=\"row_inner col_align_top tb_col_count_2 tf_box tf_rel\">\n                        <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_i23w384 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_myah864   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p><strong>Sunday, 2:45 PM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_4yfi186 last\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_g7yl209   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <h3>Parallel Computing for Two-Stage Stochastic Infrastructure Planning<\/h3>\n<p><strong>Speakers: T<span class=\"NormalTextRun SCXW139913627 BCX0\">om\u00e1s Valencia Zuluaga<\/span><span class=\"NormalTextRun SCXW139913627 BCX0\">,\u00a0<\/span><span class=\"NormalTextRun SCXW139913627 BCX0\">Elizabeth Glist<\/span><span class=\"NormalTextRun SCXW139913627 BCX0\">a,\u00a0<\/span><span class=\"NormalTextRun SCXW139913627 BCX0\">Amelia Musselman<\/span><span class=\"NormalTextRun SCXW139913627 BCX0\">,<\/span><span class=\"NormalTextRun SCXW139913627 BCX0\">\u00a0and Jean-Paul Watson<\/span><\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text --><!-- module accordion -->\n<div  class=\"module module-accordion tb_0431675 \" data-behavior=\"toggle\" data-lazy=\"1\">\n    \n    <ul class=\"ui module-accordion   tb_default_color\">\n            <li>\n            <div class=\"accordion-title tf_rel\">\n                <a href=\"#acc-0431675-0\" class=\"tb_title_accordion\" aria-controls=\"acc-0431675-0-content\" aria-expanded=\"false\">\n                    <i class=\"accordion-icon\"><svg  class=\"tf_fa tf-ti-plus\" aria-hidden=\"true\"><use href=\"#tf-ti-plus\"><\/use><\/svg><\/i>                    <i class=\"accordion-active-icon tf_hide\"><svg  class=\"tf_fa tf-ti-minus\" aria-hidden=\"true\"><use href=\"#tf-ti-minus\"><\/use><\/svg><\/i>                    <span class=\"accordion-title-wrap\">Presentation and Speaker Bios<\/span>                <\/a>\n            <\/div><!-- .accordion-title -->\n            <div id=\"acc-0431675-0-content\" data-id=\"acc-0431675-0\" aria-hidden=\"true\" class=\"accordion-content tf_hide tf_clearfix\">\n                        <div  data-lazy=\"1\" class=\"module_subrow themify_builder_sub_row tf_w col_align_top tb_col_count_1 tb_t7qt705\">\n                <div  data-lazy=\"1\" class=\"module_column sub_column tb_8cej705 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_7q7p705   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p>Infrastructure planning has become increasingly difficult in recent years as <br>natural hazards affect supply and demand patterns as well as the network of equipment connecting the two. In order to plan coordinated infrastructure systems that are resilient to a variety of potential threats, it is necessary to represent these systems at sufficiently high resolution to capture geographic and temporal variations as well as uncertainty. Both of these factors translate into much larger optimization problems than have traditionally been considered, and solving these problems is at the frontier of what is computationally feasible. Parallel computing is a tool to push that frontier. In this tutorial, we present methods for solving large-scale two-stage stochastic mixed-integer linear programming (MILP) problems using high-performance computing (HPC) resources, with a focus on infrastructure planning problems.<\/p>\n<p>To this end, we cover the necessary basics of modeling stochastic infrastructure planning problems and leveraging parallel computing resources to solve stochastic MILPs. We discuss decomposition algorithms and their parallel implementation in the Python package mpi-sppy and present examples of how to use this tool to solve stochastic infrastructure\u00a0planning problems. Finally, we present an example of how mpi-sppy has\u00a0been used to solve a realistically sized power system expansion planning problem for\u00a0California to demonstrate the difficulty of solving large-scale, stochastic, infrastructure\u00a0planning problems and how HPC resources can be leveraged to solve such problems.<\/p>\n<p><strong>Elizabeth Glista<\/strong>\u00a0is a research scientist at Lawrence Livermore National Laboratory (LLNL). Her research focuses on optimization problems with applications to power system planning, operation, and resilience, including large-scale stochastic planning and non-convex operational problems. At LLNL, she has contributed to internal research projects and to work supported by the U.S. Department of Energy\u2019s Office of Electricity through the Advanced Grid Modeling (AGM) and North American Energy Resilience Model (NAERM) projects. She holds a PhD and an MS in Mechanical Engineering, Controls, from the University of California, Berkeley (2023, 2018) and a BS in Mechanical Engineering from the Massachusetts Institute of Technology (2017). Her work has been recognized with the IEEE Power &amp; Energy Society General Meeting Best Conference Paper Award (2022) and the American Control Conference Best Student Paper Award (2020). Amelia Musselman is an operations research engineer at Lawrence Livermore National Laboratory. She also has experience at Sandia National Laboratories, RAND Corporation, and Pacific Northwest National Laboratory. She holds a Ph.D. in Industrial Engineering and M.S. in Operations Research from Georgia Institute of Technology as well as a B.S. in Mathematics from Harvey Mudd College. Her expertise is in applications of optimization, including multi-objective, stochastic, and robust optimization, to critical infrastructure planning and protection. She has experience working on problems such as stochastic unit commitment, capacity expansion planning, power system and control network restoration, distributed optimization for electric vehicle charging, and other areas of power system planning and protection under uncertainty.<\/p>\n<p><strong>Tomas Valencia Zuluaga<\/strong>\u00a0is a postdoctoral researcher in the Center for Applied Scientific Computing at Lawrence Livermore National Laboratory (LLNL). His research lies at the intersection of optimization and power systems applications, including planning, operation and electricity markets. At LLNL, he has worked on developing high-performance-computing tools for optimal power grid expansion planning under uncertainty. He obtained his PhD in Industrial Engineering and Operations Research (2024) from UC Berkeley, where he received the IEOR Faculty Fellowship award (2023) and Outstanding Graduate Student Instructor Award (2022). He also holds a MS in Electrical Engineering (2018) and BS in Mechatronics Engineering (2014) from Universidad Nacional de Colombia in Bogot\u00e1, and has professional experience in the Oil &amp; Gas and Power sectors in Colombia.<\/p>\n<p><strong>Jean-Paul Watson (&#8220;JP&#8221;)<\/strong>\u00a0is a Distinguished Member of Technical Staff in the Computational Engineering Division (CED) at Lawrence Livermore National Laboratory (LLNL), in Livermore California. At LLNL, he is the Associate Program Lead for Disaster Resilience in the Global Security Directorate. JP leads a diverse team of researchers focused on developing advanced analytics for critical infrastructure operations, planning, and resilience \u2013 emphasizing decision-making under uncertainty. He is co-inventor of the widely used Pyomo algebraic modeling language for mathematical optimization (www.pyomo.org), and has co-authored over 75 journal articles, 30 conference papers, and 3 books. JP has received the R&amp;D 100 award, the INFORMS Computing Society Prize, and was an INFORMS Edelman finalist.<\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <\/div>\n                    <\/div><!-- .accordion-content -->\n        <\/li>\n        <\/ul>\n\n<\/div><!-- \/module accordion -->        <\/div>\n                        <\/div>\n        <\/div>\n                        <div  data-lazy=\"1\" class=\"module_row themify_builder_row tb_lpys948 tf_w\">\n                        <div class=\"row_inner col_align_top tb_col_count_2 tf_box tf_rel\">\n                        <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_2pew948 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_cday948   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p><strong>Sunday, 4:15 PM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_o9ts948 last\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_6iia963   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <h3>Fraud Analytics as a Sequential Decision System: Integrating Machine Learning, Optimization, and Adversarial Learning<\/h3>\n<p><strong>Speaker: Tahir Ekin<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text --><!-- module accordion -->\n<div  class=\"module module-accordion tb_ba1x362 \" data-behavior=\"toggle\" data-lazy=\"1\">\n    \n    <ul class=\"ui module-accordion   tb_default_color\">\n            <li>\n            <div class=\"accordion-title tf_rel\">\n                <a href=\"#acc-ba1x362-0\" class=\"tb_title_accordion\" aria-controls=\"acc-ba1x362-0-content\" aria-expanded=\"false\">\n                    <i class=\"accordion-icon\"><svg  class=\"tf_fa tf-ti-plus\" aria-hidden=\"true\"><use href=\"#tf-ti-plus\"><\/use><\/svg><\/i>                    <i class=\"accordion-active-icon tf_hide\"><svg  class=\"tf_fa tf-ti-minus\" aria-hidden=\"true\"><use href=\"#tf-ti-minus\"><\/use><\/svg><\/i>                    <span class=\"accordion-title-wrap\">Presentation and Speaker Bio<\/span>                <\/a>\n            <\/div><!-- .accordion-title -->\n            <div id=\"acc-ba1x362-0-content\" data-id=\"acc-ba1x362-0\" aria-hidden=\"true\" class=\"accordion-content tf_hide tf_clearfix\">\n                        <div  data-lazy=\"1\" class=\"module_subrow themify_builder_sub_row tf_w col_align_top tb_col_count_1 tb_hqqo381\">\n                <div  data-lazy=\"1\" class=\"module_column sub_column tb_3b7u381 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_xy3m381   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p>Fraud analytics is fundamentally a decision problem under uncertainty,<br>involving trade-offs between detection performance, investigation costs, and operational constraints. While statistical and machine learning models are widely used to detect anomalous behavior and estimate fraud risk, they are often deployed without explicit consideration of downstream decisions, strategic adaptation, and system-level dynamics.<\/p>\n<p>This tutorial presents a unified framework that reframes fraud analytics as a sequential decision system involving adaptive and adversarial agents. We <br>review descriptive and predictive models, and embed them within a decision-theoretic framework that captures trade-offs between false positives, false negatives, and resource constraints. We further introduce adversarial risk analysis and related approaches to model strategic interactions between fraudsters and detection systems. By connecting prediction, optimization, and adversarial modeling, the tutorial highlights how decisions\u00a0influence both operational outcomes and future data through feedback effects.\u00a0This sequential perspective emphasizes the need for adaptive policies that account\u00a0for evolving fraud <br>behavior and changing system conditions. Examples from health\u00a0care and financial fraud illustrate how analytical models support real-world decision-making\u00a0and resource allocation in high-stakes environments. Intended for researchers\u00a0and practitioners across operations research, statistics, and data science, this tutorial\u00a0provides an accessible synthesis that requires no prior background. By the end, the\u00a0reader will understand how to frame fraud detection as a decision problem, evaluate the\u00a0limitations of purely predictive approaches, and reason about adversarial adaptation\u00a0within a unified sequential framework applicable across fraud domains.<\/p>\n<p><strong>Tahir Ekin<\/strong>\u00a0is Fields Chair in Business Analytics and Professor of Analytics at Texas State University, where he also serves as the founding Director of the Center for Analytics and Data Science (TXST CADS). His research focuses on probabilistic modeling, statistical learning, decision-making under uncertainty, adversarial machine learning, and fraud analytics, with applications in health care. He is the author of &#8220;Statistics and Health Care Fraud: How to Save Billions&#8221; (ASA\u2013CRC Press). His work has been supported by the National Science Foundation, Air Force Office of Scientific Research, and Texas Health and Human Services. He holds a Ph.D. in Decision Sciences from The George Washington University.<\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <\/div>\n                    <\/div><!-- .accordion-content -->\n        <\/li>\n        <\/ul>\n\n<\/div><!-- \/module accordion -->        <\/div>\n                        <\/div>\n        <\/div>\n                        <div  data-lazy=\"1\" class=\"module_row themify_builder_row tb_d8s8585 tf_w\">\n                        <div class=\"row_inner col_align_top tb_col_count_2 tf_box tf_rel\">\n                        <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_o97f585 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_epr8585   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p><strong>Monday, 8:00 AM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_wyqg585 last\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_r6k0155   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <h3>Multiagent Online Learning in Dynamic and <br>Uncertain Environments<\/h3>\n<p><strong>Speakers: Ceyhun Eksin, Jeff S. Shamma, Behrouz Touri<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text --><!-- module accordion -->\n<div  class=\"module module-accordion tb_h5sa493 \" data-behavior=\"toggle\" data-lazy=\"1\">\n    \n    <ul class=\"ui module-accordion   tb_default_color\">\n            <li>\n            <div class=\"accordion-title tf_rel\">\n                <a href=\"#acc-h5sa493-0\" class=\"tb_title_accordion\" aria-controls=\"acc-h5sa493-0-content\" aria-expanded=\"false\">\n                    <i class=\"accordion-icon\"><svg  class=\"tf_fa tf-ti-plus\" aria-hidden=\"true\"><use href=\"#tf-ti-plus\"><\/use><\/svg><\/i>                    <i class=\"accordion-active-icon tf_hide\"><svg  class=\"tf_fa tf-ti-minus\" aria-hidden=\"true\"><use href=\"#tf-ti-minus\"><\/use><\/svg><\/i>                    <span class=\"accordion-title-wrap\">Presentation and Speaker Bios<\/span>                <\/a>\n            <\/div><!-- .accordion-title -->\n            <div id=\"acc-h5sa493-0-content\" data-id=\"acc-h5sa493-0\" aria-hidden=\"true\" class=\"accordion-content tf_hide tf_clearfix\">\n                        <div  data-lazy=\"1\" class=\"module_subrow themify_builder_sub_row tf_w col_align_top tb_col_count_1 tb_1oii511\">\n                <div  data-lazy=\"1\" class=\"module_column sub_column tb_958d511 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_ghck511   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <div class=\"ewa-rteLine\">Multiagent online learning studies how multiple decision-making agents adapt their behavior over time in response to strategic interaction, uncertainty, and<\/div>\n<div class=\"ewa-rteLine\">non-stationary environments created by other adaptive agents. Such settings arise naturally in large-scale engineered and socio-technical systems, including transportation networks, energy markets, financial systems, supply chains, and emerging agentic AI platforms, where agents may be any combination of humans, algorithms, or physical systems.<\/div>\n<div class=\"ewa-rteLine\">\u00a0<\/div>\n<div class=\"ewa-rteLine\">This article presents a tutorial overview of learning in games and evolutionary game theory as a foundational framework for modeling and analyzing these interactions. We introduce core game-theoretic concepts and discuss how these outcomes may emerge under adaptive learning dynamics, along with selected impossibility results that capture obstacles to these outcomes. Representative discrete-time and continuous-time learning algorithms and their connections are reviewed alongside their convergence and long-run properties. The tutorial further presents learning for stochastic and Markov game settings, drawing connections to multiagent reinforcement learning and illustrating how strategic-form learning results can be leveraged in this generalized setting.<\/div>\n<div class=\"ewa-rteLine\">\u00a0<\/div>\n<div>\n<div class=\"ewa-rteLine\"><strong>Ceyhun Eksin <\/strong>is an Associate Professor with the Department of Industrial and Systems Engineering, Texas A&amp;M\u00a0University, College Station, TX, USA. He received the B.Sc. degree in Control Engineering from Istanbul Technical University, Istanbul, Turkey, in 2005, the M.S. degree in Industrial Engineering from Bogazi\u00e7i University, Istanbul, in 2008, the M.A. degree in statistics from the Wharton Statistics Department, and the Ph.D. degree in Electrical and Systems Engineering from University of Pennsylvania, in 2015. He was a Post-Doctoral Researcher jointly affiliated with the School of Biological Sciences and the School of Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, GA, USA. He was a recipient of the NSF CAREER Award in 2023. His research interests include networks, game theory, control theory, and distributed optimization.<\/div>\n<div class=\"ewa-rteLine\">\u00a0<\/div>\n<div class=\"ewa-rteLine\"><strong>Jeff Shamma<\/strong> is the Department Head of Industrial and Enterprise Systems Engineering and Jerry S. Dobrovolny Chair at the University of Illinois Urbana-Champaign. He previously held faculty positions at the King Abdullah University of Science and Technology (KAUST) and at Georgia Tech as 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 past Distinguished Lecturer of the IEEE Control Systems Society, and a recipient of the IFAC High Impact Paper Award, AACC Donald P. Eckman Award, and NSF Young Investigator Award. Jeff has been a plenary\/semi-plenary speaker at NeurIPS, World Congress of the Game Theory Society, and IEEE Conference on Decision and Control. He was Editor-in-Chief of the IEEE Transactions on Control of Network Systems from 2020-2024. Jeff\u2019s research focuses on decision and control, game theory, and multi-agent systems.<\/div>\n<div class=\"ewa-rteLine\">\u00a0<\/div>\n<div class=\"ewa-rteLine\"><strong>Behrouz Touri <\/strong>is an Associate Professor of the Industrial and Enterprise Systems Engineering at the University of Illinois at Urbana-Champaign (UIUC) and an affiliate faculty of the ECE Departments at UIUC and University of California, San Diego (UCSD). Prior to joining UIUC, he was an Associate Professor of ECE at the UCSD. He received his B.Sc. degree in Electrical Engineering from Isfahan University of Technology, Isfahan, Iran in 2006, his M.Sc. degree in Communications, Systems, Electronics from Jacobs University, Bremen, Germany in 2008, and his Ph.D. degree in Industrial Engineering from University of Illinois at Urbana-Champaign in 2011. His research interests include applied probability theory, distributed optimization, control and estimation, population dynamics, and game\u00a0theory.<\/div>\n<\/div>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <\/div>\n                    <\/div><!-- .accordion-content -->\n        <\/li>\n        <\/ul>\n\n<\/div><!-- \/module accordion -->        <\/div>\n                        <\/div>\n        <\/div>\n                        <div  data-lazy=\"1\" class=\"module_row themify_builder_row tb_ok4w415 tf_w\">\n                        <div class=\"row_inner col_align_top tb_col_count_2 tf_box tf_rel\">\n                        <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_qqlt415 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_0p3p415   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p><strong>Monday, 11:00 AM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_wf90415 last\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_m091101   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <h3>Multi-objective Combinatorial Optimization: Foundations, Theory, and Methods<\/h3>\n<p><strong>Speaker: Banu Lokman<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text --><!-- module accordion -->\n<div  class=\"module module-accordion tb_x6k5389 \" data-behavior=\"toggle\" data-lazy=\"1\">\n    \n    <ul class=\"ui module-accordion   tb_default_color\">\n            <li>\n            <div class=\"accordion-title tf_rel\">\n                <a href=\"#acc-x6k5389-0\" class=\"tb_title_accordion\" aria-controls=\"acc-x6k5389-0-content\" aria-expanded=\"false\">\n                    <i class=\"accordion-icon\"><svg  class=\"tf_fa tf-ti-plus\" aria-hidden=\"true\"><use href=\"#tf-ti-plus\"><\/use><\/svg><\/i>                    <i class=\"accordion-active-icon tf_hide\"><svg  class=\"tf_fa tf-ti-minus\" aria-hidden=\"true\"><use href=\"#tf-ti-minus\"><\/use><\/svg><\/i>                    <span class=\"accordion-title-wrap\">Presentation and Speaker Bio<\/span>                <\/a>\n            <\/div><!-- .accordion-title -->\n            <div id=\"acc-x6k5389-0-content\" data-id=\"acc-x6k5389-0\" aria-hidden=\"true\" class=\"accordion-content tf_hide tf_clearfix\">\n                        <div  data-lazy=\"1\" class=\"module_subrow themify_builder_sub_row tf_w col_align_top tb_col_count_1 tb_h2h8411\">\n                <div  data-lazy=\"1\" class=\"module_column sub_column tb_9n4d411 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_uwol411   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p>This tutorial presents the foundations, theory and methods of multi-objective<br>integer programming, with a particular focus on multi-objective combinatorial<br>optimization (MOCO). Many real-world optimization problems are inherently<br>combinatorial in nature and involve multiple, often conflicting, objectives. MOCO has been widely used to model these complex decision problems across various domains such as transportation, logistics, finance, energy, and healthcare. While the feasible set is typically finite in MOCO problems, these solutions are not always explicitly available to the decision-makers. Adding to this complexity, there is rarely a single solution that optimizes all objectives simultaneously in multi-objective optimization problems.<\/p>\n<p>This tutorial provides a structured framework for students, researchers, and practitioners to approach and solve these complex decision problems. We first define key terminology, describe the main characteristics of MOCO problems and discuss their scalarization. We then explore the advanced methods and exact algorithms to generate all nondominated points, for which an improvement in one objective cannot be made without sacrificing performance in another. Since these algorithms become\u00a0intractable in real-world problem settings with the increase in the number of nondominated\u00a0points, we also discuss the methods that generate a representative set of\u00a0solutions with a prespecified level of quality or find preferred solutions. The tutorial\u00a0provides a broad and accessible overview of existing methods, supported by illustrative\u00a0examples, discussions, figures, and comprehensive references that clarify their\u00a0main ideas, strengths, and limitations.<\/p>\n<p><strong>Banu Lokman<\/strong> is a Professor of Operational Research in the School of Organisations, Systems and People at the University of Portsmouth, United Kingdom. She completed her PhD in Industrial Engineering at Middle East Technical University (METU) in Turkey and held academic and research positions at METU and the Aalto University School of Business in Finland. She currently serves as the President of the INFORMS Section on Multiple Criteria Decision Making (MCDM), having previously held board positions in the International Society on MCDM and the INFORMS MCDM Section. She is an Associate Editor for Omega and the IMA Journal of Management Mathematics and is an editorial board member of the Journal of Multi-Criteria Decision Analysis. She is also a member of the Research Committee of the UK Operational Research Society and leads UK-based MCDM courses at NATCOR. In 2022, she received the Bernard Roy Award from the Association of European Operational Research Societies (EURO) Working Group on Multiple Criteria Decision Aiding for her contributions to the field. Her research interests include MCDM, combinatorial optimization, multi-objective integer and mixed-integer programming, clustering, and applications in energy, sustainability, and healthcare. Her work has been published in several peer-reviewed journals, including Management Science and European Journal of Operational Research.<\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <\/div>\n                    <\/div><!-- .accordion-content -->\n        <\/li>\n        <\/ul>\n\n<\/div><!-- \/module accordion -->        <\/div>\n                        <\/div>\n        <\/div>\n                        <div  data-lazy=\"1\" class=\"module_row themify_builder_row tb_bx6i627 tf_w\">\n                        <div class=\"row_inner col_align_top tb_col_count_2 tf_box tf_rel\">\n                        <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_dgwl627 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_q55t627   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p><strong>Monday, 1:15 PM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_lkhr627 last\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_if73141   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <h3>Self-Adapting Approximations of Markov Decision Processes: A Guided Tour<\/h3>\n<p><strong>Speakers: Andre Augusto Cire, Selvaprabu Nadarajah, Parshan Pakiman, Negar Soheili<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text --><!-- module accordion -->\n<div  class=\"module module-accordion tb_w4si858 \" data-behavior=\"toggle\" data-lazy=\"1\">\n    \n    <ul class=\"ui module-accordion   tb_default_color\">\n            <li>\n            <div class=\"accordion-title tf_rel\">\n                <a href=\"#acc-w4si858-0\" class=\"tb_title_accordion\" aria-controls=\"acc-w4si858-0-content\" aria-expanded=\"false\">\n                    <i class=\"accordion-icon\"><svg  class=\"tf_fa tf-ti-plus\" aria-hidden=\"true\"><use href=\"#tf-ti-plus\"><\/use><\/svg><\/i>                    <i class=\"accordion-active-icon tf_hide\"><svg  class=\"tf_fa tf-ti-minus\" aria-hidden=\"true\"><use href=\"#tf-ti-minus\"><\/use><\/svg><\/i>                    <span class=\"accordion-title-wrap\">Presentation and Speaker Bios<\/span>                <\/a>\n            <\/div><!-- .accordion-title -->\n            <div id=\"acc-w4si858-0-content\" data-id=\"acc-w4si858-0\" aria-hidden=\"true\" class=\"accordion-content tf_hide tf_clearfix\">\n                        <div  data-lazy=\"1\" class=\"module_subrow themify_builder_sub_row tf_w col_align_top tb_col_count_1 tb_e8ms878\">\n                <div  data-lazy=\"1\" class=\"module_column sub_column tb_c6zn878 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_ti3r878   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p>Sequential decision making under uncertainty arises across business, <br>engineering, and science. Markov Decision Processes (MDPs) offer a rich modeling language, but translating models into implementable policies requires confronting challenging state and action spaces and non-convex optimization landscapes. Approximations are essential, and making them accessible and performant without algorithmic expertise allows users to focus on formulating MDPs that capture real-world features rather than\u00a0simplifying their models to facilitate solution. Obtaining a policy involves a three-stage COR cycle: Constructing a parametrized model that approximates the MDP, Optimizing for the parameters of this model, and Refining the model by learning from the\u00a0solution. These stages involve design choices that burden the user, leading them to simplify or skip stages.<\/p>\n<p>This tutorial provides an overview of methods that reduce the user burden using COR as its guide. Construction ranges from lightweight extensible models to automated structural reformulations. Optimization leverages mathematical programming and first-order methods that have matured in theory and off-the-shelf software. Refinement performs feature expansion and approximation tightening, feeding back into the construct stage. When a combination of techniques across COR stages results in both accessibility and performance, we refer to it as a self-adapting approximation. Such an approximation adapts to problem structure, instance data, or solution progress. We guide readers through these ideas across weakly coupled MDPs and general continuous state-action MDPs, providing generalizable foundations and a code base to get started on a simple example. The tutorial concludes by outlining the potential for foundation models and quantum computing to further impact this area.<\/p>\n<p><strong>Andre A. Cire<\/strong> is an Associate Professor of Operations Management and Analytics at the University of Toronto, cross-appointed between the Rotman School of Management and the Department of Management at the Scarborough campus. His research bridges methodological development and practical applications in optimization, with a focus on sequential decision-making, mathematical programming, dynamic programming, network models, and applications in scheduling, healthcare, and supply chains. Andre earned his PhD in Operations Research from Carnegie Mellon University, and his contributions have been recognized with awards such as the INFORMS Harvey J. Greenberg Award (2024), an honorable mention for the INFORMS Computing Society Prize (2023), and research and teaching awards at the University of Toronto. He has also held senior editorial and service roles for multiple journals and conferences including INFORMS Journal on Computing, AAAI, and NeurIPS.<\/p>\n<p><strong>Selvaprabu (Selva) Nadarajah<\/strong> is an Associate Professor of Information and Decision Sciences and Bielinski Family Endowed Scholar at the College of Business Administration, University of Illinois Chicago (UIC). He is also a UIC Global Scholar and previously served as the Decision Intelligence R&amp;D Lead at the Discovery Partners Institute, the innovation hub of the University of Illinois System. He is an Associate Editor for Decision Sciences, Operations Research, and Production and Operations Management. His research develops self-adapting approximation methods for large-scale Markov decision processes. He applies these methods to energy investment and operations problems, including real options valuation of commodity and energy assets, renewable energy procurement, and capacity expansion. This work has been recognized with awards from INFORMS, the Commodity and Energy Markets Association, and NeurIPS. It has been funded by the Alfred P. Sloan Foundation, Argonne National Laboratory, and industry. He has received the UIC College of Business Teaching Excellence Award and the EnergyTech University Prize Faculty Explorer Award from the U.S. Department of Energy for educational initiatives. He holds a PhD in Operations Research from the Tepper School of Business at Carnegie Mellon University.<\/p>\n<p><strong>Parshan Pakiman<\/strong> is an Assistant Professor in the Department of Operations Management and Strategy at the University at Buffalo School of Management. His research focuses on approximate dynamic programming and data-driven optimization methods for sequential decision making, with an emphasis on large-scale Markov decision processes and structured relaxations. His work develops algorithms with strong theoretical guarantees that minimize the need for human intervention during deployment, including model selection and parameter tuning. These methods are designed to be scalable, interpretable, and accessible to practitioners, with applications in healthcare operations and revenue management. Prior to joining the University at Buffalo, he was a Principal Researcher at the Tolan Center for Healthcare at the University of Chicago Booth School of Business. He received his PhD in Information and Decision Sciences from the University of Illinois Chicago and his BSc in Applied Mathematics from the University of Tehran.<\/p>\n<p><strong>Negar Soheili<\/strong> is an Associate Professor in the Department of Information and Decision Sciences at the University of Illinois Chicago. She received her PhD in Operations Research from Carnegie Mellon University. Her research develops scalable optimization algorithms for machine learning and sequential decision-making under uncertainty. In particular, she designs efficient first-order methods for large-scale constrained optimization and the approximation of Markov decision processes, with applications in healthcare, supply chains, and business analytics. She launched the PhD program in Operations Research and Business Analytics at UIC. She serves as an Associate Editor for the INFORMS Journal on Computing and for Optimization Letters.<\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <\/div>\n                    <\/div><!-- .accordion-content -->\n        <\/li>\n        <\/ul>\n\n<\/div><!-- \/module accordion -->        <\/div>\n                        <\/div>\n        <\/div>\n                        <div  data-lazy=\"1\" class=\"module_row themify_builder_row tb_pwc1391 tf_w\">\n                        <div class=\"row_inner col_align_top tb_col_count_2 tf_box tf_rel\">\n                        <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_64s3391 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_a3we391   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p><strong>Monday, 2:45 PM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_5f8m391 last\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_oh1a983   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <h3>Decision-Focused Learning: When and Why Traditional Prediction Models Fail<\/h3>\n<p><strong>Speaker: Mo Liu<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text --><!-- module accordion -->\n<div  class=\"module module-accordion tb_mag8105 \" data-behavior=\"toggle\" data-lazy=\"1\">\n    \n    <ul class=\"ui module-accordion   tb_default_color\">\n            <li>\n            <div class=\"accordion-title tf_rel\">\n                <a href=\"#acc-mag8105-0\" class=\"tb_title_accordion\" aria-controls=\"acc-mag8105-0-content\" aria-expanded=\"false\">\n                    <i class=\"accordion-icon\"><svg  class=\"tf_fa tf-ti-plus\" aria-hidden=\"true\"><use href=\"#tf-ti-plus\"><\/use><\/svg><\/i>                    <i class=\"accordion-active-icon tf_hide\"><svg  class=\"tf_fa tf-ti-minus\" aria-hidden=\"true\"><use href=\"#tf-ti-minus\"><\/use><\/svg><\/i>                    <span class=\"accordion-title-wrap\">Presentation and Speaker Bio<\/span>                <\/a>\n            <\/div><!-- .accordion-title -->\n            <div id=\"acc-mag8105-0-content\" data-id=\"acc-mag8105-0\" aria-hidden=\"true\" class=\"accordion-content tf_hide tf_clearfix\">\n                        <div  data-lazy=\"1\" class=\"module_subrow themify_builder_sub_row tf_w col_align_top tb_col_count_1 tb_63jo123\">\n                <div  data-lazy=\"1\" class=\"module_column sub_column tb_h8el123 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_2a35123   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p>Plugging predictions of unknown parameters into downstream optimization<br>problems, often referred to as the \u201cpredict-then-optimize\u201d paradigm, has long<br>been a standard approach in decision-making under uncertainty. However, improved\u00a0predictive accuracy does not, in general, translate into improved decision quality.\u00a0This disconnect has motivated growing interest in decision-focused learning (DFL)\u00a0within the operations research community.<\/p>\n<p>This tutorial reviews recent developments\u00a0in DFL and highlights key methodological insights, with a particular focus on stochastic\u00a0linear programming as the downstream decision-making problem. We discuss why\u00a0several widely used tools in traditional statistical learning are not directly suited to\u00a0decision-focused settings and must be rethought, including (i) data collection strategies\u00a0driven purely by predictive uncertainty and (ii) distributional distance measures\u00a0such as the Wasserstein distance. We summarize properties of DFL that distinguish it\u00a0from conventional predictive modeling and provide insights into the development of\u00a0new decision-focused tools.<\/p>\n<p><strong>Mo Liu<\/strong> is an assistant professor in the Department of Statistics and Operations Research at the University of North Carolina at Chapel Hill. He received his PhD in Industrial Engineering and Operations Research from the University of California, Berkeley. His research centers on decision-focused learning, a methodology that designs and trains prediction models while accounting for downstream optimization problems. These downstream problems arise in real-world applications such as revenue and inventory management.<\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <\/div>\n                    <\/div><!-- .accordion-content -->\n        <\/li>\n        <\/ul>\n\n<\/div><!-- \/module accordion -->        <\/div>\n                        <\/div>\n        <\/div>\n                        <div  data-lazy=\"1\" class=\"module_row themify_builder_row tb_0amo758 tf_w\">\n                        <div class=\"row_inner col_align_top tb_col_count_2 tf_box tf_rel\">\n                        <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_cyi7758 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_wmxc758   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p><strong>Monday, 4:15 PM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_q1d8758 last\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_f60q506   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <h3>Generative Models for Decision-Making under Distributional Shift<\/h3>\n<p><strong>Speakers: Xiuyuan Cheng, Yunqin Zhu, Yao Xie<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text --><!-- module accordion -->\n<div  class=\"module module-accordion tb_kt0u63 \" data-behavior=\"toggle\" data-lazy=\"1\">\n    \n    <ul class=\"ui module-accordion   tb_default_color\">\n            <li>\n            <div class=\"accordion-title tf_rel\">\n                <a href=\"#acc-kt0u63-0\" class=\"tb_title_accordion\" aria-controls=\"acc-kt0u63-0-content\" aria-expanded=\"false\">\n                    <i class=\"accordion-icon\"><svg  class=\"tf_fa tf-ti-plus\" aria-hidden=\"true\"><use href=\"#tf-ti-plus\"><\/use><\/svg><\/i>                    <i class=\"accordion-active-icon tf_hide\"><svg  class=\"tf_fa tf-ti-minus\" aria-hidden=\"true\"><use href=\"#tf-ti-minus\"><\/use><\/svg><\/i>                    <span class=\"accordion-title-wrap\">Presentation and Speaker Bios<\/span>                <\/a>\n            <\/div><!-- .accordion-title -->\n            <div id=\"acc-kt0u63-0-content\" data-id=\"acc-kt0u63-0\" aria-hidden=\"true\" class=\"accordion-content tf_hide tf_clearfix\">\n                        <div  data-lazy=\"1\" class=\"module_subrow themify_builder_sub_row tf_w col_align_top tb_col_count_1 tb_19r583\">\n                <div  data-lazy=\"1\" class=\"module_column sub_column tb_ybsd83 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_mcdw83   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <div class=\"ewa-rteLine\">Many data-driven decision problems are formulated using a nominal distribution<br>estimated from historical data, while performance is ultimately determined by<br>a deployment distribution that may be shifted, context-dependent, partially observed,\u00a0or stress-induced. This tutorial presents modern generative models, particularly flow and\u00a0score-based methods, as mathematical tools for constructing decision-relevant\u00a0distributions. From an operations research perspective, their primary value lies not\u00a0in unconstrained sample synthesis but in representing and transforming distributions\u00a0through transport maps, velocity fields, score fields, and guided stochastic dynamics.\u00a0We present a unified framework based on pushforward maps, continuity, Fokker\u2013\u00a0Planck equations, Wasserstein geometry, and optimization in probability space.<\/div>\n<div>\u00a0<\/div>\n<div class=\"ewa-rteLine\">Within this framework, generative models can be used to learn nominal uncertainty, construct stressed or least-favorable distributions for robustness, and produce conditional or\u00a0posterior distributions under side information and partial observation. We also highlight\u00a0representative theoretical guarantees, including forward\u2013reverse convergence for\u00a0iterative flow models, first-order minimax analysis in transport-map space, and error transfer\u00a0bounds for posterior sampling with generative priors. The tutorial provides\u00a0a principled introduction to using generative models for scenario generation, robust\u00a0decision-making, uncertainty quantification, and related problems under distributional\u00a0shift.<\/div>\n<div class=\"ewa-rteLine\">\u00a0<\/div>\n<div>\n<p><span style=\"font-weight: bold\">Xiuyuan Cheng<\/span> is a professor of mathematics at Duke University. She received her Ph.D. degree from Princeton University in 2013. Before joining Duke, she was a postdoctoral researcher at \u00c9cole Normale Sup\u00e9rieure in Paris from 2013 to 2015 and a Gibbs Assistant Professor at Yale University from 2015 to 2017.\u00a0Her research interests include theoretical and computational techniques for high-dimensional data\u00a0analysis, signal processing, and machine learning. She is a recipient of a Sloan Fellowship and an\u00a0NSF CAREER Award.<\/p>\n<p><strong>Yunqin \u201cVinci\u201d Zhu<\/strong> is a Ph.D. student in the H. Milton Stewart School of Industrial and Systems Engineering at the Georgia Institute of Technology. He received his B.Eng. degree in artificial intelligence from the University of Science and Technology (USTC) of China. His research interests include machine learning, generative modeling, and optimization. He is a recipient of the Guo Moruo Scholarship from USTC.<\/p>\n<p><strong>Yao Xie<\/strong>\u00a0 is the Coca-Cola\u00a0Foundation Chair and Professor in the H. Milton Stewart School of Industrial and Systems Engineering at the Georgia Institute of Technology, where she is also Associate Director of the Machine Learning Center. She received her Ph.D. degree in electrical engineering, with a minor in mathematics, from Stanford University and was previously a Research Scientist at Duke University.\u00a0Her research lies at the intersection of statistics, machine learning, and optimization, with a focus\u00a0on developing statistically powerful and computationally efficient methods for high-dimensional,\u00a0sequential, and spatio-temporal data. She is a Member of Cohort 2026 of the National Academies\u2019\u00a0New Voices in Sciences, Engineering, and Medicine program and the IEEE Information Theory\u00a0Society Distinguished Lecturer for 2026\u20132027. Her honors include the NSF CAREER Award,\u00a0INFORMS Wagner Prize Finalist, INFORMS Gaver Early Career Award, and C.W.S. Woodroofe\u00a0Award. She serves as an associate editor for several journals, including IEEE Transactions on\u00a0Information Theory, Journal of the American Statistical Association\u2014Theory and Methods, Operations\u00a0Research, Annals of Applied Statistics, Sequential Analysis, and INFORMS Journal on Data\u00a0Science.<\/p>\n<\/div>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <\/div>\n                    <\/div><!-- .accordion-content -->\n        <\/li>\n        <\/ul>\n\n<\/div><!-- \/module accordion -->        <\/div>\n                        <\/div>\n        <\/div>\n                        <div  data-lazy=\"1\" class=\"module_row themify_builder_row tb_n00l539 tf_w\">\n                        <div class=\"row_inner col_align_top tb_col_count_2 tf_box tf_rel\">\n                        <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_l6oq539 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_n5pw966   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p><strong>Tuesday, 11:00 AM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_8ekw802 last\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_9e8t249   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <h3>GPU-Accelerated Decision Optimization<\/h3>\n<p><strong>Speakers: Nicolas Blin, Burcin Bozkaya, Akif \u00c7\u00f6rd\u00fck, Chris Maes<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text --><!-- module accordion -->\n<div  class=\"module module-accordion tb_jwql938 \" data-behavior=\"toggle\" data-lazy=\"1\">\n    \n    <ul class=\"ui module-accordion   tb_default_color\">\n            <li>\n            <div class=\"accordion-title tf_rel\">\n                <a href=\"#acc-jwql938-0\" class=\"tb_title_accordion\" aria-controls=\"acc-jwql938-0-content\" aria-expanded=\"false\">\n                    <i class=\"accordion-icon\"><svg  class=\"tf_fa tf-ti-plus\" aria-hidden=\"true\"><use href=\"#tf-ti-plus\"><\/use><\/svg><\/i>                    <i class=\"accordion-active-icon tf_hide\"><svg  class=\"tf_fa tf-ti-minus\" aria-hidden=\"true\"><use href=\"#tf-ti-minus\"><\/use><\/svg><\/i>                    <span class=\"accordion-title-wrap\">Presentation and Speaker Bios<\/span>                <\/a>\n            <\/div><!-- .accordion-title -->\n            <div id=\"acc-jwql938-0-content\" data-id=\"acc-jwql938-0\" aria-hidden=\"true\" class=\"accordion-content tf_hide tf_clearfix\">\n                        <div  data-lazy=\"1\" class=\"module_subrow themify_builder_sub_row tf_w col_align_top tb_col_count_1 tb_wusv961\">\n                <div  data-lazy=\"1\" class=\"module_column sub_column tb_hg8t961 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_0q2t961   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p>For decades, algorithms for solving decision optimization problems have been designed and implemented for CPUs.\u00a0Recent advances in GPU hardware, driven by machine learning and AI, together with the development of GPU-accelerated scientific computing\u00a0kernels, have made GPUs a practical platform for solving optimization problems. This shift is most visible in linear\u00a0programming, where first-order methods exploit GPU parallelism and high-bandwidth memory to solve problems with millions\u00a0of variables and constraints. GPU algorithms for mixed-integer programming are still emerging, while GPU-based vehicle-routing<br>solvers are more mature and already support large-scale routing applications. At the same time, generative AI and agentic AI\u00a0are changing how optimization models are formulated, solved, and embedded in decision workflows.<\/p>\n<p>This tutorial introduces\u00a0GPU-accelerated decision optimization for researchers and practitioners familiar with CPU-based modeling and solvers. We review<br>computational patterns that make optimization algorithms amenable to GPU acceleration, summarize the growing ecosystem of opensource\u00a0and commercial GPU solvers, and provide benchmarks that quantify current performance. We also present cuOpt, NVIDIA\u2019s\u00a0open-source library for GPU-accelerated decision optimization. Our goal is to help the optimization community understand where\u00a0GPUs are useful today, where important limitations remain, and where future research and development are needed.<\/p>\n<p><strong>Nicolas Blin<\/strong>\u00a0is a senior developer technology engineer at NVIDIA. Nicolas holds an M.Sc. in software engineering. He has a background in image processing and GPU-accelerated algorithms. His interests include massively parallel algorithms, performance optimization, and combinatorial and linear optimization.<\/p>\n<p><strong>Burcin Bozkaya\u00a0<\/strong>is a senior developer relations manager at NVIDIA. He holds a BS and MS in Industrial Engineering and PhD in Management Science, specializing in combinatorial optimization problems and heuristic optimization as applied in transportation and logistics, and supply chain planning. As the developer relations lead for decision optimization, Burcin engages with the OR developer ecosystem to evangelize and support the open-source community, ISV partners and key business planners for developing accelerated optimization solvers.<\/p>\n<p><strong>Akif Coerduek<\/strong>\u00a0is a senior developer technology engineer at NVIDIA. Akif holds a bachelor\u2019s degree in computer engineering and a master\u2019s degree in software engineering. He has a background in massively parallel pricing engines and GPU-accelerated combinatorial optimization. His interests include parallel algorithms, performance optimization, and combinatorial optimization.<\/p>\n<p><strong>Christopher Maes<\/strong>\u00a0is a principal engineer on the NVIDIA cuOpt team. He received his PhD in Computational and Mathematical Engineering from Stanford University in 2010. His background and research interests are in linear, quadratic, and mixed-integer programming, as well as trajectory optimization, and machine learning.<\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <\/div>\n                    <\/div><!-- .accordion-content -->\n        <\/li>\n        <\/ul>\n\n<\/div><!-- \/module accordion -->        <\/div>\n                        <\/div>\n        <\/div>\n                        <div  data-lazy=\"1\" class=\"module_row themify_builder_row tb_n5gg68 tf_w\">\n                        <div class=\"row_inner col_align_top tb_col_count_2 tf_box tf_rel\">\n                        <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_zmpn68 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_8bvb68   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p><strong>Tuesday, 1:15 PM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_x2he68 last\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_6oyl177   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <h3>A Modern Treatment of the Primal\u2013Dual Framework for Online Resource Allocation<\/h3>\n<p><strong>Speakers: Rad Niazadeh, Rajan Udwani<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text --><!-- module accordion -->\n<div  class=\"module module-accordion tb_nz4v225 \" data-behavior=\"toggle\" data-lazy=\"1\">\n    \n    <ul class=\"ui module-accordion   tb_default_color\">\n            <li>\n            <div class=\"accordion-title tf_rel\">\n                <a href=\"#acc-nz4v225-0\" class=\"tb_title_accordion\" aria-controls=\"acc-nz4v225-0-content\" aria-expanded=\"false\">\n                    <i class=\"accordion-icon\"><svg  class=\"tf_fa tf-ti-plus\" aria-hidden=\"true\"><use href=\"#tf-ti-plus\"><\/use><\/svg><\/i>                    <i class=\"accordion-active-icon tf_hide\"><svg  class=\"tf_fa tf-ti-minus\" aria-hidden=\"true\"><use href=\"#tf-ti-minus\"><\/use><\/svg><\/i>                    <span class=\"accordion-title-wrap\">Presentation and Speaker Bios<\/span>                <\/a>\n            <\/div><!-- .accordion-title -->\n            <div id=\"acc-nz4v225-0-content\" data-id=\"acc-nz4v225-0\" aria-hidden=\"true\" class=\"accordion-content tf_hide tf_clearfix\">\n                        <div  data-lazy=\"1\" class=\"module_subrow themify_builder_sub_row tf_w col_align_top tb_col_count_1 tb_8y4q246\">\n                <div  data-lazy=\"1\" class=\"module_column sub_column tb_dmem246 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_3mpp246   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p>Linear-programming (LP)-based primal\u2013dual methods are fundamental for designing and analyzing algorithms in adversarial (prior-free) online resource allocation. This chapter provides a tutorial on two modern primal-dual frameworks, emphasizing recent developments and contemporary models in operations research.<\/p>\n<p>Part I develops an LP-based convex-programming framework where solving a regularized convex program at each arrival captures the tradeoff between greediness and hedging, yielding a dual certificate via Karush\u2013Kuhn\u2013Tucker\u00a0(KKT) conditions. Because standard LP relaxations can be weak or intractable\u00a0for stochastic outcomes, Part II introduces a complementary LP-free framework that provides a universal certificate system for evaluating competitive ratios\u00a0under such uncertainty. Covering a wide array of models\u2014including online vertex-weighted bipartite matching, edge-weighted online matching with free disposal, online matching with stochastic rewards, reusable resources, two-sided assortment optimization, configuration allocation (whole-page optimization), AdWords, and costly cancellations\u2014the tutorial equips readers with versatile <br>proof templates to analyze existing algorithms and develop new solutions for emerging applications.<\/p>\n<div class=\"ewa-rteLine\"><strong>Rad Niazadeh<\/strong> is an Associate Professor of Operations Management at the University of Chicago Booth School of Business. He is also part of the faculty at Toyota Technological Institute of Chicago (TTIC) by a courtesy appointment, and a faculty advisor at Lyft Inc. (Fulfillment &amp; Matching Science). Prior to joining Chicago Booth, he was a visiting researcher at Google Research NYC and a Motwani postdoctoral fellow at Stanford University, Computer Science Department. He obtained his PhD in Computer Science (minored in Applied Mathematics) from Cornell University. His research spans online algorithms, online learning, and mechanism design, with applications in theory and practice of online platforms and non-profit operations. His work has received multiple recognitions including the INFORMS Auctions and Market Design Rothkopf Junior Researcher Paper Award (2021, 2024: first place; 2023: second place), the INFORMS Revenue Management and Pricing Dissertation Award (honorable mention), and best student paper awards at INFORMS George Nicholson Competition (2025: honorable mention), INFORMS MSOM (2024), INFORMS Service Science (2025: finalist), INFORMS Data Mining, and IJCAI conference.<\/div>\n<div>\u00a0<\/div>\n<div class=\"ewa-rteLine\"><strong>Rajan Udwani<\/strong> is an Assistant Professor of Industrial Engineering and Operations Research at the University of California, Berkeley. His research focuses on algorithms for optimization under uncertainty, with a particular emphasis on revenue management and pricing. Before joining UC Berkeley, he was a postdoctoral researcher at Columbia University. He holds a Ph.D. in Operations Research from MIT and a B.Tech. in Electrical Engineering from IIT Bombay. <br>His work has been recognized with INFORMS junior faculty and student paper awards, and is supported by an NSF CAREER Award and the Google Research Scholar Program.<\/div>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <\/div>\n                    <\/div><!-- .accordion-content -->\n        <\/li>\n        <\/ul>\n\n<\/div><!-- \/module accordion -->        <\/div>\n                        <\/div>\n        <\/div>\n                        <div  data-lazy=\"1\" class=\"module_row themify_builder_row tb_39d8752 tf_w\">\n                        <div class=\"row_inner col_align_top tb_col_count_2 tf_box tf_rel\">\n                        <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_nhx1752 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_bvq3752   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p><strong>Tuesday, 2:45 PM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <div  data-lazy=\"1\" class=\"module_column tb-column col4-2 tb_0jpb752 last\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_0zs2529   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <h3>Foundations of Reinforcement Learning and Control: Connections and New Perspective<\/h3>\n<p><strong>Speakers: Claire Vernade, Onno Eberhard Max, Martha White, Florian D\u00f6rfler, Csaba Szepevari, Miroslav Krstic, Michael Muehlebach<\/strong><\/p>    <\/div>\n<\/div>\n<!-- \/module text --><!-- module accordion -->\n<div  class=\"module module-accordion tb_gvs9752 \" data-behavior=\"toggle\" data-lazy=\"1\">\n    \n    <ul class=\"ui module-accordion   tb_default_color\">\n            <li>\n            <div class=\"accordion-title tf_rel\">\n                <a href=\"#acc-gvs9752-0\" class=\"tb_title_accordion\" aria-controls=\"acc-gvs9752-0-content\" aria-expanded=\"false\">\n                    <i class=\"accordion-icon\"><svg  class=\"tf_fa tf-ti-plus\" aria-hidden=\"true\"><use href=\"#tf-ti-plus\"><\/use><\/svg><\/i>                    <i class=\"accordion-active-icon tf_hide\"><svg  class=\"tf_fa tf-ti-minus\" aria-hidden=\"true\"><use href=\"#tf-ti-minus\"><\/use><\/svg><\/i>                    <span class=\"accordion-title-wrap\">Presentation and Speaker Bios<\/span>                <\/a>\n            <\/div><!-- .accordion-title -->\n            <div id=\"acc-gvs9752-0-content\" data-id=\"acc-gvs9752-0\" aria-hidden=\"true\" class=\"accordion-content tf_hide tf_clearfix\">\n                        <div  data-lazy=\"1\" class=\"module_subrow themify_builder_sub_row tf_w col_align_top tb_col_count_1 tb_qsbm772\">\n                <div  data-lazy=\"1\" class=\"module_column sub_column tb_9w8c772 first\">\n                    <!-- module text -->\n<div  class=\"module module-text tb_mpeq772   \" data-lazy=\"1\">\n        <div  class=\"tb_text_wrap\">\n        <p>Reinforcement learning and control theory are two adjacent scientific fields<br>that focus on optimizing the controller of unknown dynamical systems using feedback.\u00a0While both fields have common roots in dynamic programming, they have evolved with\u00a0distinct methodologies, goals, and cultures. Despite decades of mutual influence, a\u00a0significant gap persists between the two communities.<\/p>\n<p>This tutorial introduces adaptive\u00a0control, actor-critic reinforcement algorithms, and a new original way to combine\u00a0these two paradigms for data-driven decision making on a classical locomotion control\u00a0problem. Our aim is to provide keys to understand the core differences between both\u00a0approaches, and insights to help experts in each field better understand and engage\u00a0with the tools and approaches of the other.<\/p>\n<div class=\"ewa-rteLine\"><strong>Claire Vernade<\/strong> is Full Professor of Foundations of Machine Learning at the University of Technology Nuremberg (UTN). Her research focuses on sequential decision making under uncertainty, spanning reinforcement learning, online learning, and statistical machine learning. She develops theoretically grounded learning and decision-making algorithms for adaptive and interactive systems, with an emphasis on bridging mathematical foundations and scalable machine learning methods. Before joining UTN in 2025, she was a Group Leader at the University of T\u00fcbingen and a Senior Research Scientist at Google DeepMind. She is the recipient of an Emmy Noether Programme grant and an ERC Starting Grant.<\/div>\n<div class=\"ewa-rteLine\"><br><strong>Onno Eberhard<\/strong> is a Ph.D. student in Computer Science at the Max Planck <br>Institute for Intelligent Systems and the University of T\u00fcbingen. He holds an M.Sc. in Machine Learning from the University of T\u00fcbingen and a B.Sc. in Electrical Engineering from the University of Duisburg-Essen, and has gained professional experience at Google Research and Siemens.\u00a0His research focuses on the theoretical foundations of reinforcement learning, particularly regarding <br>partially observable environments, recurrent memory, and its intersections with control theory.<\/div>\n<div class=\"ewa-rteLine\"><br><strong>Martha White<\/strong> is a Professor of Computing Science at the University of Alberta and a Fellow of Amii, which is one of the top machine learning centres in the world. She holds a Canada CIFAR AI Chair, a Tier 2 Canada Research Chair in Reinforcement Learning, received IEEE\u2019s \u201cAIs 10 to Watch: The Future of AI\u201d award in 2020 and was inducted into the College of New Scholars by the Royal Society of Canada in 2024. She has authored more than 80 papers in top journals and conferences. Martha is an associate editor for JMLR and TMLR, server on the RLC board and has served as co-program chair for ICLR and for RLC. Her research focus is on developing reinforcement learning algorithms that learn to adapt continually, with a focus on process control and more sustainable systems.<\/div>\n<div class=\"ewa-rteLine\"><br><strong>Florian D\u00f6rfler<\/strong> is a Professor at the Automatic Control Laboratory at ETH Z\u00fcrich. He received his Ph.D. degree in Mechanical Engineering from the University of California at Santa Barbara in 2013. From 2013 to 2014 he was an Assistant Professor at the University of California Los Angeles. His research interests are centered around automatic control, system theory, optimization, and learning. His particular foci are on network systems, data-driven settings, and applications to power systems. He is a recipient of the R\u00f6ssler Prize, the highest scientific award at ETH Z\u00fcrich across all disciplines, as well as the distinguished career awards by IFAC (Manfred Thoma Medal) and EUCA (European Control Award). He and his team received best paper distinctions in the top venues of control, power systems, power electronics, circuits and systems.<\/div>\n<div class=\"ewa-rteLine\"><br><strong>Csaba Szepesv\u00e1ri<\/strong> is a Canada CIFAR AI Chair, Professor of Computing Science <br>at the University of Alberta, and Team Lead for the Foundations team at DeepMind. He is a Fellow of the Association for the Advancement of Artificial Intelligence and an IEEE Fellow. He received his PhD in 1999 from J\u00f3zsef Attila University in Szeged, Hungary, in probability and statistics. His research <br>advances the foundations of learning-based artificial intelligence, especially reinforcement learning, bandit algorithms, and planning under uncertainty. He is the author\u00a0or co-author of three books, including Bandit Algorithms, published by Cambridge University Press in 2020. He is also the co-inventor of UCT, an algorithm that helped ignite the modern development of Monte Carlo tree search and made simulation-based planning a central tool in game AI and decision making under uncertainty.<\/div>\n<div class=\"ewa-rteLine\"><br><strong>Miroslav Krstic<\/strong> is a professor and serves as senior associate vice chancellor for research at UC San Diego. He is the recipient of the IEEE Brockett Award and Bode Prize, ASME Oldenburger Medal, SIAM Reid Prize, Bellman Award, and other recognitions, including the Chestnut prize and several IFAC TC awards. He is a member of the Serbian Academy of Sciences and Arts, Academia Europaea, fellow of IEEE, IFAC, SIAM, ASME, AIAA, and other societies, and Fellow-Ambassador of CNRS. Krstic is the current editor-in-chief of IEEE Transactions on Automatic Control, a former EiC of Systems &amp; Control Letters, and former senior editor in Automatica. He is a coauthor of 19 books and several hundred papers on various nonlinear, adaptive, and infinite-dimensional control subjects.<\/div>\n<div class=\"ewa-rteLine\"><br><strong>Michael M\u00fchlebach<\/strong> leads the research group learning and dynamical systems at the Max Planck Institute for Intelligent Systems in T\u00fcbingen, Germany. His group conducts fundamental research in machine learning, reinforcement learning, and large-scale optimization. He won numerous awards including an Emmy Noether and Branco Weiss fellowship, as well as an ETH Medal and the HILTI prize for innovative research. He is also a member of the editorial board of Foundations and Trends in Machine Learning.<\/div>    <\/div>\n<\/div>\n<!-- \/module text -->        <\/div>\n                    <\/div>\n                    <\/div><!-- .accordion-content -->\n        <\/li>\n        <\/ul>\n\n<\/div><!-- \/module accordion -->        <\/div>\n                        <\/div>\n        <\/div>\n        <\/div>\n<!--\/themify_builder_content-->","protected":false},"excerpt":{"rendered":"<p>TutORials<\/p>\n","protected":false},"author":46,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"_acf_changed":false,"content-type":"","footnotes":""},"class_list":["post-12076","page","type-page","status-publish","hentry","has-post-title","has-post-date","has-post-category","has-post-tag","has-post-comment","has-post-author",""],"acf":[],"yoast_head":"<!-- This site is optimized with the Yoast SEO Premium plugin v26.0 (Yoast SEO v26.0) - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>TutORials &#187; 2026 INFORMS Annual Meeting<\/title>\n<meta name=\"description\" content=\"The TutORials in Operations Research series is published annually by INFORMS as an introduction to emerging and classical subfields of operations research and management science.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/meetings.informs.org\/wordpress\/annual\/tutorials\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"TutORials\" \/>\n<meta property=\"og:description\" content=\"The TutORials in Operations Research series is published annually by INFORMS as an introduction to emerging and classical subfields of operations research and management science.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/meetings.informs.org\/wordpress\/annual\/tutorials\/\" \/>\n<meta property=\"og:site_name\" content=\"2026 INFORMS Annual Meeting\" \/>\n<meta property=\"article:modified_time\" content=\"2026-08-17T11:38:58+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/meetings.informs.org\/wordpress\/annual\/files\/2025\/10\/2026_INFORMS_Annual_Meeting_Logo.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"300\" \/>\n\t<meta property=\"og:image:height\" content=\"300\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data1\" content=\"94 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/meetings.informs.org\/wordpress\/annual\/tutorials\/\",\"url\":\"https:\/\/meetings.informs.org\/wordpress\/annual\/tutorials\/\",\"name\":\"TutORials &#187; 2026 INFORMS Annual Meeting\",\"isPartOf\":{\"@id\":\"https:\/\/meetings.informs.org\/wordpress\/annual\/#website\"},\"datePublished\":\"2026-07-20T15:17:13+00:00\",\"dateModified\":\"2026-08-17T11:38:58+00:00\",\"description\":\"The TutORials in Operations Research series is published annually by INFORMS as an introduction to emerging and classical subfields of operations research and management science.\",\"breadcrumb\":{\"@id\":\"https:\/\/meetings.informs.org\/wordpress\/annual\/tutorials\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/meetings.informs.org\/wordpress\/annual\/tutorials\/\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/meetings.informs.org\/wordpress\/annual\/tutorials\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\/\/meetings.informs.org\/wordpress\/annual\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"TutORials\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/meetings.informs.org\/wordpress\/annual\/#website\",\"url\":\"https:\/\/meetings.informs.org\/wordpress\/annual\/\",\"name\":\"2026 INFORMS Annual Meeting\",\"description\":\"November 1-4, 2026 | San Francisco, CA\",\"publisher\":{\"@id\":\"https:\/\/meetings.informs.org\/wordpress\/annual\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/meetings.informs.org\/wordpress\/annual\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Organization\",\"@id\":\"https:\/\/meetings.informs.org\/wordpress\/annual\/#organization\",\"name\":\"2026 INFORMS Annual Meeting\",\"url\":\"https:\/\/meetings.informs.org\/wordpress\/annual\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\/\/meetings.informs.org\/wordpress\/annual\/#\/schema\/logo\/image\/\",\"url\":\"https:\/\/meetings.informs.org\/wordpress\/annual\/files\/2025\/10\/cropped-2026_INFORMS_Annual_Meeting_Logo.png\",\"contentUrl\":\"https:\/\/meetings.informs.org\/wordpress\/annual\/files\/2025\/10\/cropped-2026_INFORMS_Annual_Meeting_Logo.png\",\"width\":512,\"height\":512,\"caption\":\"2026 INFORMS Annual Meeting\"},\"image\":{\"@id\":\"https:\/\/meetings.informs.org\/wordpress\/annual\/#\/schema\/logo\/image\/\"}}]}<\/script>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"TutORials &#187; 2026 INFORMS Annual Meeting","description":"The TutORials in Operations Research series is published annually by INFORMS as an introduction to emerging and classical subfields of operations research and management science.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/meetings.informs.org\/wordpress\/annual\/tutorials\/","og_locale":"en_US","og_type":"article","og_title":"TutORials","og_description":"The TutORials in Operations Research series is published annually by INFORMS as an introduction to emerging and classical subfields of operations research and management science.","og_url":"https:\/\/meetings.informs.org\/wordpress\/annual\/tutorials\/","og_site_name":"2026 INFORMS Annual Meeting","article_modified_time":"2026-08-17T11:38:58+00:00","og_image":[{"width":300,"height":300,"url":"https:\/\/meetings.informs.org\/wordpress\/annual\/files\/2025\/10\/2026_INFORMS_Annual_Meeting_Logo.jpg","type":"image\/jpeg"}],"twitter_card":"summary_large_image","twitter_misc":{"Est. reading time":"94 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/meetings.informs.org\/wordpress\/annual\/tutorials\/","url":"https:\/\/meetings.informs.org\/wordpress\/annual\/tutorials\/","name":"TutORials &#187; 2026 INFORMS Annual Meeting","isPartOf":{"@id":"https:\/\/meetings.informs.org\/wordpress\/annual\/#website"},"datePublished":"2026-07-20T15:17:13+00:00","dateModified":"2026-08-17T11:38:58+00:00","description":"The TutORials in Operations Research series is published annually by INFORMS as an introduction to emerging and classical subfields of operations research and management science.","breadcrumb":{"@id":"https:\/\/meetings.informs.org\/wordpress\/annual\/tutorials\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/meetings.informs.org\/wordpress\/annual\/tutorials\/"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/meetings.informs.org\/wordpress\/annual\/tutorials\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/meetings.informs.org\/wordpress\/annual\/"},{"@type":"ListItem","position":2,"name":"TutORials"}]},{"@type":"WebSite","@id":"https:\/\/meetings.informs.org\/wordpress\/annual\/#website","url":"https:\/\/meetings.informs.org\/wordpress\/annual\/","name":"2026 INFORMS Annual Meeting","description":"November 1-4, 2026 | San Francisco, CA","publisher":{"@id":"https:\/\/meetings.informs.org\/wordpress\/annual\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/meetings.informs.org\/wordpress\/annual\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Organization","@id":"https:\/\/meetings.informs.org\/wordpress\/annual\/#organization","name":"2026 INFORMS Annual Meeting","url":"https:\/\/meetings.informs.org\/wordpress\/annual\/","logo":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/meetings.informs.org\/wordpress\/annual\/#\/schema\/logo\/image\/","url":"https:\/\/meetings.informs.org\/wordpress\/annual\/files\/2025\/10\/cropped-2026_INFORMS_Annual_Meeting_Logo.png","contentUrl":"https:\/\/meetings.informs.org\/wordpress\/annual\/files\/2025\/10\/cropped-2026_INFORMS_Annual_Meeting_Logo.png","width":512,"height":512,"caption":"2026 INFORMS Annual Meeting"},"image":{"@id":"https:\/\/meetings.informs.org\/wordpress\/annual\/#\/schema\/logo\/image\/"}}]}},"builder_content":"<h1>TutORials<\/h1>\n<p>The <em>TutORials in Operations Research<\/em> series is published annually by INFORMS as an introduction to emerging and classical subfields of operations research and management science. These chapters are designed to be accessible for all constituents of the INFORMS community, including current students, practitioners, faculty, and researchers. The publication allows readers to keep pace with new developments in the field and serves as augmenting material for a selection of the tutorial presentations offered at the INFORMS Annual Meeting.<\/p>\n<p><strong>Sunday, 8:00 AM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>\n<h3>Transform Method for Stochastic Processing and Matching Networks<\/h3> <p><strong>Speakers: Sushil\u202fVarma, Prakirt Jhunjhunwala, Daniela Hurtado-Lange, Siva <br>Theja Maguluri<\/strong><\/p>\n<ul><li><h4>Presentation and Speaker Bios<\/h4>Modern service systems\u2014ranging from cloud data centers and ride-hailing platforms to healthcare facilities\u2014operate at massive scales where congestion is a critical challenge. Utilizing an operations research approach, these systems are analyzed by modeling them as complex stochastic processes, which are typically understood through process-level convergence to fluid and diffusion limits. However, these methods often prove technically dense and provide limited guidance for finite, practical system scales. \u00a0 The transform method, presented in this tutorial, was recently developed as a unified and tractable framework for the steady-state analysis of Stochastic Processing and Matching Networks (SPNs\/SMNs). The transform method overcomes the technical hurdles\u00a0of process-level convergence by working directly with the pre-limit system. By exploiting the zero-drift property of exponential test functions, the method derives explicit functional equations (acting as a proxy for global balance equations) for the transforms (such as moment-generating functions) of queue-length distributions. This approach provides sharp, non-asymptotic performance guarantees, bridging the gap between theoretical asymptotics and real-world system behavior. Since its introduction in 2020 for load-balancing in data center networks, the transform method has been extended to handle realistic complexities, including customer abandonment, state-dependent arrivals, Markov-modulated arrivals, large-system scale, and multi-dimensional networks with multiple bottlenecks. We survey these theoretical advances and demonstrate their practical relevance across diverse domains, such as matching markets and networked service systems. The transform method provides interpretable bounds tied directly to system parameters, offering a powerful analytical alternative to simulation-heavy or purely asymptotic approaches for system design and control. \u00a0 <strong>Sushil Mahavir Varma<\/strong> is an assistant professor in the Industrial and Operations Engineering Department at the University of Michigan, Ann Arbor. Before joining Michigan, he was a postdoctoral researcher at INRIA Paris. He received his PhD degree in operations research from Georgia Institute of Technology. His research lies broadly in applied probability. He has worked on problems spanning two-sided matching markets, electric vehicle operations, queueing theory, load balancing, and graph alignment. His work has been recognized with the 2024 ACM SIGMETRICS Doctoral Dissertation Award, the 2025 Georgia Tech Sigma Xi <br>Best PhD Thesis Award, and finalist recognition for the 2025 INFORMS TSL Dissertation Award. \u00a0 <strong>Siva Theja Magulur<\/strong>i is Fouts Family Early Career Professor and Associate Professor in the H. Milton Stewart School of Industrial and Systems Engineering at Georgia Tech. He received his B.Tech in Electrical Engineering from IIT Madras, M.S in ECE, M.S. in Applied Math and a PhD in ECE all from University of Illinois at Urbana Champaign. His research interests span the areas of Networks, Control, Optimization, Algorithms, Applied Probability and Reinforcement Learning. He is a recipient of the biennial \u201cBest Publication in Applied Probability\u201d award, NSF CAREER award, \u201cCTL\/BP Junior Faculty Teaching Excellence Award,\u201d and \u201cStudent Recognition of Excellence in Teaching: Class of 1934 CIOS Award.\u201d \u00a0 <strong>Prakirt Jhunjhunwala<\/strong> is a Postdoctoral Scientist at Amazon in the FBA Science team. Previously, he was a Postdoctoral Research Scholar at Columbia Business School, New York. He received his Ph.D. in Operations Research and a Masters in Mathematics from Georgia Tech in 2023 and his B.Tech. (Honors) in Electrical Engineering from IIT Bombay. His research focuses on the design and analysis of stochastic networks, with applications in data centers and quantum networks. Prakirt received an Honorable Mention for the SIGMETRICS Doctoral Dissertation Award (2023), the Best Paper Award at SPCOM (2018), and the Ed Iacobucci Fellowship for Excellence in Applied Probability and Simulation (2022). \u00a0 <strong>Daniela Hurtado-Lange<\/strong> is an assistant professor in the Operations Department at the Kellogg School of Management at Northwestern University. Before joining Kellogg, she was an assistant professor of Mathematics at William &amp; Mary for 1.5 years. She obtained her Ph.D. in Operations Research from Georgia Tech in December 2021, and her research focuses on performance analysis of stochastic processing networks. Specifically, she works on heavy-traffic analysis and queueing theory. Her work has been recognized with the 2022 Sigma Xi Best Ph.D. Thesis Award, and second place in the 2020 JFIG competition.<\/li><\/ul>\nModern service systems\u2014ranging from cloud data centers and ride-hailing platforms to healthcare facilities\u2014operate at massive scales where congestion is a critical challenge. Utilizing an operations research approach, these systems are analyzed by modeling them as complex stochastic processes, which are typically understood through process-level convergence to fluid and diffusion limits. However, these methods often prove technically dense and provide limited guidance for finite, practical system scales. \u00a0 The transform method, presented in this tutorial, was recently developed as a unified and tractable framework for the steady-state analysis of Stochastic Processing and Matching Networks (SPNs\/SMNs). The transform method overcomes the technical hurdles\u00a0of process-level convergence by working directly with the pre-limit system. By exploiting the zero-drift property of exponential test functions, the method derives explicit functional equations (acting as a proxy for global balance equations) for the transforms (such as moment-generating functions) of queue-length distributions. This approach provides sharp, non-asymptotic performance guarantees, bridging the gap between theoretical asymptotics and real-world system behavior. Since its introduction in 2020 for load-balancing in data center networks, the transform method has been extended to handle realistic complexities, including customer abandonment, state-dependent arrivals, Markov-modulated arrivals, large-system scale, and multi-dimensional networks with multiple bottlenecks. We survey these theoretical advances and demonstrate their practical relevance across diverse domains, such as matching markets and networked service systems. The transform method provides interpretable bounds tied directly to system parameters, offering a powerful analytical alternative to simulation-heavy or purely asymptotic approaches for system design and control. \u00a0\n<strong>Sushil Mahavir Varma<\/strong> is an assistant professor in the Industrial and Operations Engineering Department at the University of Michigan, Ann Arbor. Before joining Michigan, he was a postdoctoral researcher at INRIA Paris. He received his PhD degree in operations research from Georgia Institute of Technology. His research lies broadly in applied probability. He has worked on problems spanning two-sided matching markets, electric vehicle operations, queueing theory, load balancing, and graph alignment. His work has been recognized with the 2024 ACM SIGMETRICS Doctoral Dissertation Award, the 2025 Georgia Tech Sigma Xi <br>Best PhD Thesis Award, and finalist recognition for the 2025 INFORMS TSL Dissertation Award. \u00a0 <strong>Siva Theja Magulur<\/strong>i is Fouts Family Early Career Professor and Associate Professor in the H. Milton Stewart School of Industrial and Systems Engineering at Georgia Tech. He received his B.Tech in Electrical Engineering from IIT Madras, M.S in ECE, M.S. in Applied Math and a PhD in ECE all from University of Illinois at Urbana Champaign. His research interests span the areas of Networks, Control, Optimization, Algorithms, Applied Probability and Reinforcement Learning. He is a recipient of the biennial \u201cBest Publication in Applied Probability\u201d award, NSF CAREER award, \u201cCTL\/BP Junior Faculty Teaching Excellence Award,\u201d and \u201cStudent Recognition of Excellence in Teaching: Class of 1934 CIOS Award.\u201d \u00a0 <strong>Prakirt Jhunjhunwala<\/strong> is a Postdoctoral Scientist at Amazon in the FBA Science team. Previously, he was a Postdoctoral Research Scholar at Columbia Business School, New York. He received his Ph.D. in Operations Research and a Masters in Mathematics from Georgia Tech in 2023 and his B.Tech. (Honors) in Electrical Engineering from IIT Bombay. His research focuses on the design and analysis of stochastic networks, with applications in data centers and quantum networks. Prakirt received an Honorable Mention for the SIGMETRICS Doctoral Dissertation Award (2023), the Best Paper Award at SPCOM (2018), and the Ed Iacobucci Fellowship for Excellence in Applied Probability and Simulation (2022). \u00a0 <strong>Daniela Hurtado-Lange<\/strong> is an assistant professor in the Operations Department at the Kellogg School of Management at Northwestern University. Before joining Kellogg, she was an assistant professor of Mathematics at William &amp; Mary for 1.5 years. She obtained her Ph.D. in Operations Research from Georgia Tech in December 2021, and her research focuses on performance analysis of stochastic processing networks. Specifically, she works on heavy-traffic analysis and queueing theory. Her work has been recognized with the 2022 Sigma Xi Best Ph.D. Thesis Award, and second place in the 2020 JFIG competition.\n<p><strong>Sunday, 11:00 AM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>\n<h3>Deep Learning for Sequential Decision Making under Uncertainty: Foundations, Frameworks, and Frontiers<\/h3> <p><strong>Speaker: Esra B\u00fcy\u00fcktahtak\u0131n Toy<\/strong><\/p>\n<ul><li><h4>Presentation and Speaker Bio<\/h4><p>Artificial intelligence (AI) is moving beyond prediction toward systems supporting decisions in complex, dynamic environments. This shift creates a natural<br>intersection with operations research and management science (OR\/MS), which<br>has long provided methodological foundations for sequential decision making under uncertainty. At the same time, deep learning advances, including feedforward neural networks, recurrent architectures, transformers, large language models (LLMs), and deep reinforcement learning, have expanded data-driven modeling for large-scale decisions.<\/p> <p>This tutorial presents an OR\/MS-centered perspective on deep learning for sequential decision making under uncertainty, bridging neural architectures and OR\/MS approaches to decision making. Its premise: deep learning complements optimization rather than replacing it. Deep learning brings adaptability and scalable approximation, whereas OR\/MS provides the mathematical rigor to represent constraints, recourse, uncertainty, and decision quality. The tutorial reviews key decision making foundations, connects them to the major neural architectures in modern AI, and organizes the field around three central themes: predict-then-optimize and decision-aware\u00a0learning, learning-based decision generation under constraints for continuous\u00a0and discrete problems with temporal coupling, and deep reinforcement learning for\u00a0sequential and combinatorial decision making. Impact spans supply chains, service\u00a0systems, healthcare and epidemic response, agriculture, energy, environmental sustainability,\u00a0and autonomous operations. This tutorial frames these developments as\u00a0part of a shift from predictive AI toward decision-capable AI, highlighting OR\/MS\u2019s\u00a0role in shaping the next generation of integrated learning\u2013optimization systems.<\/p> <p><strong>Esra B\u00fcy\u00fcktahtak\u0131n<\/strong>\u00a0is a Full Professor of Operations Research in the Grado Department of Industrial and Systems Engineering at Virginia Tech, where she directs the Systems\u00a0Optimization and Machine Learning Lab (SysOptiMaL). She earned her Ph.D. in Operations\u00a0Research from the University of Florida. Her research advances multi-stage stochastic mixed-integer\u00a0programming through optimization theory, algorithmic innovation, and learning-enhanced\u00a0optimization. She is recognized for pioneering contributions at the interface of deep learning and\u00a0optimization for sequential decision making under uncertainty. Her work bridges operations research,\u00a0machine learning, and artificial intelligence, with applications in health systems, epidemic\u00a0supply chains, biosecurity, defense, ecological conservation, agriculture, forestry, and environmental\u00a0sustainability.<\/p> <p>Dr. B\u00fcy\u00fcktahtak\u0131n is the recipient of the National Science Foundation (NSF) CAREER Award (2016) and the INFORMS Minority Issues Forum (MIF) Early Career Award (2016), and her research\u00a0has been supported by NSF, the U.S. Department of Agriculture (USDA), the Office of Naval Research (ONR), the U.S. Forest Service, the Virginia Department of Forestry, the Minnesota\u00a0Aquatic Invasive Species Research Center (MAISRC), and the 4-VA Collaborative Research\u00a0Program, with more than $3 million in external funding. She has authored 49 peer-reviewed journal\u00a0publications in leading scholarly outlets, including 27 papers in A* and A ranked journals. Her\u00a0work has received six INFORMS Best Publication Awards and has been featured three times in\u00a0ISE Magazine, as well as in the INFORMS Computing Society and U.S. Forest Service newsletters.\u00a0Her professional service includes leadership in the INFORMS community, where she served\u00a0as President of the INFORMS Junior Faculty Interest Group (JFIG, 2014\u20132015) and received two\u00a0INFORMS Service <br>Awards. She currently serves as an Associate Editor of the INFORMS Journal\u00a0<br>on Computing.<\/p><\/li><\/ul>\n<p>Artificial intelligence (AI) is moving beyond prediction toward systems supporting decisions in complex, dynamic environments. This shift creates a natural<br>intersection with operations research and management science (OR\/MS), which<br>has long provided methodological foundations for sequential decision making under uncertainty. At the same time, deep learning advances, including feedforward neural networks, recurrent architectures, transformers, large language models (LLMs), and deep reinforcement learning, have expanded data-driven modeling for large-scale decisions.<\/p> <p>This tutorial presents an OR\/MS-centered perspective on deep learning for sequential decision making under uncertainty, bridging neural architectures and OR\/MS approaches to decision making. Its premise: deep learning complements optimization rather than replacing it. Deep learning brings adaptability and scalable approximation, whereas OR\/MS provides the mathematical rigor to represent constraints, recourse, uncertainty, and decision quality. The tutorial reviews key decision making foundations, connects them to the major neural architectures in modern AI, and organizes the field around three central themes: predict-then-optimize and decision-aware\u00a0learning, learning-based decision generation under constraints for continuous\u00a0and discrete problems with temporal coupling, and deep reinforcement learning for\u00a0sequential and combinatorial decision making. Impact spans supply chains, service\u00a0systems, healthcare and epidemic response, agriculture, energy, environmental sustainability,\u00a0and autonomous operations. This tutorial frames these developments as\u00a0part of a shift from predictive AI toward decision-capable AI, highlighting OR\/MS\u2019s\u00a0role in shaping the next generation of integrated learning\u2013optimization systems.<\/p> <p><strong>Esra B\u00fcy\u00fcktahtak\u0131n<\/strong>\u00a0is a Full Professor of Operations Research in the Grado Department of Industrial and Systems Engineering at Virginia Tech, where she directs the Systems\u00a0Optimization and Machine Learning Lab (SysOptiMaL). She earned her Ph.D. in Operations\u00a0Research from the University of Florida. Her research advances multi-stage stochastic mixed-integer\u00a0programming through optimization theory, algorithmic innovation, and learning-enhanced\u00a0optimization. She is recognized for pioneering contributions at the interface of deep learning and\u00a0optimization for sequential decision making under uncertainty. Her work bridges operations research,\u00a0machine learning, and artificial intelligence, with applications in health systems, epidemic\u00a0supply chains, biosecurity, defense, ecological conservation, agriculture, forestry, and environmental\u00a0sustainability.<\/p> <p>Dr. B\u00fcy\u00fcktahtak\u0131n is the recipient of the National Science Foundation (NSF) CAREER Award (2016) and the INFORMS Minority Issues Forum (MIF) Early Career Award (2016), and her research\u00a0has been supported by NSF, the U.S. Department of Agriculture (USDA), the Office of Naval Research (ONR), the U.S. Forest Service, the Virginia Department of Forestry, the Minnesota\u00a0Aquatic Invasive Species Research Center (MAISRC), and the 4-VA Collaborative Research\u00a0Program, with more than $3 million in external funding. She has authored 49 peer-reviewed journal\u00a0publications in leading scholarly outlets, including 27 papers in A* and A ranked journals. Her\u00a0work has received six INFORMS Best Publication Awards and has been featured three times in\u00a0ISE Magazine, as well as in the INFORMS Computing Society and U.S. Forest Service newsletters.\u00a0Her professional service includes leadership in the INFORMS community, where she served\u00a0as President of the INFORMS Junior Faculty Interest Group (JFIG, 2014\u20132015) and received two\u00a0INFORMS Service <br>Awards. She currently serves as an Associate Editor of the INFORMS Journal\u00a0<br>on Computing.<\/p>\n<p><strong>Sunday, 1:15 PM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>\n<h3>A Tutorial on Reinforcement Learning for LLMs: RLHF and Beyond<\/h3> <p><strong>Speaker: Daniel Jiang<\/strong><\/p>\n<ul><li><h4>Presentation and Speaker Bio<\/h4><p>Large language models (LLMs) require post-training, which takes a foundation model and trains it to follow instructions, behave safely, and perform well across downstream use cases. Reinforcement learning (RL) post-training has become one of the\u00a0most common post-training approaches, most prominently reinforcement learning from human feedback (RLHF), where the reward reflects human preference judgments. In this tutorial, we formulate RL post-training as a sequential decision problem (a Markov decision process), covering both single-turn settings, where the model produces one response to a prompt, and multi-turn settings,\u00a0where it interacts with a user or environment over multiple rounds. We also distinguish reward signals by how directly they can be measured or inferred (verifiable, observable, or latent). We then discuss reward overoptimization, a phenomenon in which the policy exploits errors in a learned reward model, motivating a KL-regularized objective that keeps the policy close to a reference model while still optimizing reward. From this objective, we give precise step-by-step derivations of four central policy-optimization algorithms (DPO, REINFORCE, PPO, and GRPO), explicitly distinguishing exact derivations from practical approximations and heuristics.<\/p> <p>The tutorial also covers deployment paradigms, with a focus on batch-online deployment, in which a policy\u2019s interaction\u00a0data is periodically collected and used for offline retraining and subsequent redeployment. For multi-turn settings, we cover an\u00a0extension of GRPO to full trajectories collected from a training environment, and a batch-online method that performs approximate\u00a0policy iteration from logged trajectories. Finally, we discuss open research directions in RL post-training that may benefit from a\u00a0broad range of research perspectives.<\/p> <p><strong>Daniel R. Jiang<\/strong> is a Research Scientist at Meta and an Adjunct Professor of Industrial Engineering at the University of Pittsburgh. His research spans reinforcement learning, sequential decision-making, and adaptive experimentation, with a recent emphasis on reinforcement learning for LLM post-training. His work contributed to the first real-world deployment of an RL-trained language model for generative advertising on Facebook and to the training of conversational agents through multi-turn RLHF. He has also developed multi-step lookahead methods for adaptive experimentation and co-created BoTorch, a widely used open-source library for Bayesian optimization in PyTorch. His publications have appeared in journals including Management Science, Operations Research, JMLR, and Mathematics of Operations Research, and at conferences such as NeurIPS, ICML, and ICLR. Daniel received a Ph.D. in Operations Research and Financial Engineering from Princeton University and dual B.S. degrees in Electrical and Computer Engineering and Mathematics from Purdue University.<\/p><\/li><\/ul>\n<p>Large language models (LLMs) require post-training, which takes a foundation model and trains it to follow instructions, behave safely, and perform well across downstream use cases. Reinforcement learning (RL) post-training has become one of the\u00a0most common post-training approaches, most prominently reinforcement learning from human feedback (RLHF), where the reward reflects human preference judgments. In this tutorial, we formulate RL post-training as a sequential decision problem (a Markov decision process), covering both single-turn settings, where the model produces one response to a prompt, and multi-turn settings,\u00a0where it interacts with a user or environment over multiple rounds. We also distinguish reward signals by how directly they can be measured or inferred (verifiable, observable, or latent). We then discuss reward overoptimization, a phenomenon in which the policy exploits errors in a learned reward model, motivating a KL-regularized objective that keeps the policy close to a reference model while still optimizing reward. From this objective, we give precise step-by-step derivations of four central policy-optimization algorithms (DPO, REINFORCE, PPO, and GRPO), explicitly distinguishing exact derivations from practical approximations and heuristics.<\/p> <p>The tutorial also covers deployment paradigms, with a focus on batch-online deployment, in which a policy\u2019s interaction\u00a0data is periodically collected and used for offline retraining and subsequent redeployment. For multi-turn settings, we cover an\u00a0extension of GRPO to full trajectories collected from a training environment, and a batch-online method that performs approximate\u00a0policy iteration from logged trajectories. Finally, we discuss open research directions in RL post-training that may benefit from a\u00a0broad range of research perspectives.<\/p> <p><strong>Daniel R. Jiang<\/strong> is a Research Scientist at Meta and an Adjunct Professor of Industrial Engineering at the University of Pittsburgh. His research spans reinforcement learning, sequential decision-making, and adaptive experimentation, with a recent emphasis on reinforcement learning for LLM post-training. His work contributed to the first real-world deployment of an RL-trained language model for generative advertising on Facebook and to the training of conversational agents through multi-turn RLHF. He has also developed multi-step lookahead methods for adaptive experimentation and co-created BoTorch, a widely used open-source library for Bayesian optimization in PyTorch. His publications have appeared in journals including Management Science, Operations Research, JMLR, and Mathematics of Operations Research, and at conferences such as NeurIPS, ICML, and ICLR. Daniel received a Ph.D. in Operations Research and Financial Engineering from Princeton University and dual B.S. degrees in Electrical and Computer Engineering and Mathematics from Purdue University.<\/p>\n<p><strong>Sunday, 2:45 PM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>\n<h3>Parallel Computing for Two-Stage Stochastic Infrastructure Planning<\/h3> <p><strong>Speakers: Tom\u00e1s Valencia Zuluaga,\u00a0Elizabeth Glista,\u00a0Amelia Musselman,\u00a0and Jean-Paul Watson<\/strong><\/p>\n<ul><li><h4>Presentation and Speaker Bios<\/h4><p>Infrastructure planning has become increasingly difficult in recent years as <br>natural hazards affect supply and demand patterns as well as the network of equipment connecting the two. In order to plan coordinated infrastructure systems that are resilient to a variety of potential threats, it is necessary to represent these systems at sufficiently high resolution to capture geographic and temporal variations as well as uncertainty. Both of these factors translate into much larger optimization problems than have traditionally been considered, and solving these problems is at the frontier of what is computationally feasible. Parallel computing is a tool to push that frontier. In this tutorial, we present methods for solving large-scale two-stage stochastic mixed-integer linear programming (MILP) problems using high-performance computing (HPC) resources, with a focus on infrastructure planning problems.<\/p> <p>To this end, we cover the necessary basics of modeling stochastic infrastructure planning problems and leveraging parallel computing resources to solve stochastic MILPs. We discuss decomposition algorithms and their parallel implementation in the Python package mpi-sppy and present examples of how to use this tool to solve stochastic infrastructure\u00a0planning problems. Finally, we present an example of how mpi-sppy has\u00a0been used to solve a realistically sized power system expansion planning problem for\u00a0California to demonstrate the difficulty of solving large-scale, stochastic, infrastructure\u00a0planning problems and how HPC resources can be leveraged to solve such problems.<\/p> <p><strong>Elizabeth Glista<\/strong>\u00a0is a research scientist at Lawrence Livermore National Laboratory (LLNL). Her research focuses on optimization problems with applications to power system planning, operation, and resilience, including large-scale stochastic planning and non-convex operational problems. At LLNL, she has contributed to internal research projects and to work supported by the U.S. Department of Energy\u2019s Office of Electricity through the Advanced Grid Modeling (AGM) and North American Energy Resilience Model (NAERM) projects. She holds a PhD and an MS in Mechanical Engineering, Controls, from the University of California, Berkeley (2023, 2018) and a BS in Mechanical Engineering from the Massachusetts Institute of Technology (2017). Her work has been recognized with the IEEE Power &amp; Energy Society General Meeting Best Conference Paper Award (2022) and the American Control Conference Best Student Paper Award (2020). Amelia Musselman is an operations research engineer at Lawrence Livermore National Laboratory. She also has experience at Sandia National Laboratories, RAND Corporation, and Pacific Northwest National Laboratory. She holds a Ph.D. in Industrial Engineering and M.S. in Operations Research from Georgia Institute of Technology as well as a B.S. in Mathematics from Harvey Mudd College. Her expertise is in applications of optimization, including multi-objective, stochastic, and robust optimization, to critical infrastructure planning and protection. She has experience working on problems such as stochastic unit commitment, capacity expansion planning, power system and control network restoration, distributed optimization for electric vehicle charging, and other areas of power system planning and protection under uncertainty.<\/p> <p><strong>Tomas Valencia Zuluaga<\/strong>\u00a0is a postdoctoral researcher in the Center for Applied Scientific Computing at Lawrence Livermore National Laboratory (LLNL). His research lies at the intersection of optimization and power systems applications, including planning, operation and electricity markets. At LLNL, he has worked on developing high-performance-computing tools for optimal power grid expansion planning under uncertainty. He obtained his PhD in Industrial Engineering and Operations Research (2024) from UC Berkeley, where he received the IEOR Faculty Fellowship award (2023) and Outstanding Graduate Student Instructor Award (2022). He also holds a MS in Electrical Engineering (2018) and BS in Mechatronics Engineering (2014) from Universidad Nacional de Colombia in Bogot\u00e1, and has professional experience in the Oil &amp; Gas and Power sectors in Colombia.<\/p> <p><strong>Jean-Paul Watson (\"JP\")<\/strong>\u00a0is a Distinguished Member of Technical Staff in the Computational Engineering Division (CED) at Lawrence Livermore National Laboratory (LLNL), in Livermore California. At LLNL, he is the Associate Program Lead for Disaster Resilience in the Global Security Directorate. JP leads a diverse team of researchers focused on developing advanced analytics for critical infrastructure operations, planning, and resilience \u2013 emphasizing decision-making under uncertainty. He is co-inventor of the widely used Pyomo algebraic modeling language for mathematical optimization (www.pyomo.org), and has co-authored over 75 journal articles, 30 conference papers, and 3 books. JP has received the R&amp;D 100 award, the INFORMS Computing Society Prize, and was an INFORMS Edelman finalist.<\/p><\/li><\/ul>\n<p>Infrastructure planning has become increasingly difficult in recent years as <br>natural hazards affect supply and demand patterns as well as the network of equipment connecting the two. In order to plan coordinated infrastructure systems that are resilient to a variety of potential threats, it is necessary to represent these systems at sufficiently high resolution to capture geographic and temporal variations as well as uncertainty. Both of these factors translate into much larger optimization problems than have traditionally been considered, and solving these problems is at the frontier of what is computationally feasible. Parallel computing is a tool to push that frontier. In this tutorial, we present methods for solving large-scale two-stage stochastic mixed-integer linear programming (MILP) problems using high-performance computing (HPC) resources, with a focus on infrastructure planning problems.<\/p> <p>To this end, we cover the necessary basics of modeling stochastic infrastructure planning problems and leveraging parallel computing resources to solve stochastic MILPs. We discuss decomposition algorithms and their parallel implementation in the Python package mpi-sppy and present examples of how to use this tool to solve stochastic infrastructure\u00a0planning problems. Finally, we present an example of how mpi-sppy has\u00a0been used to solve a realistically sized power system expansion planning problem for\u00a0California to demonstrate the difficulty of solving large-scale, stochastic, infrastructure\u00a0planning problems and how HPC resources can be leveraged to solve such problems.<\/p> <p><strong>Elizabeth Glista<\/strong>\u00a0is a research scientist at Lawrence Livermore National Laboratory (LLNL). Her research focuses on optimization problems with applications to power system planning, operation, and resilience, including large-scale stochastic planning and non-convex operational problems. At LLNL, she has contributed to internal research projects and to work supported by the U.S. Department of Energy\u2019s Office of Electricity through the Advanced Grid Modeling (AGM) and North American Energy Resilience Model (NAERM) projects. She holds a PhD and an MS in Mechanical Engineering, Controls, from the University of California, Berkeley (2023, 2018) and a BS in Mechanical Engineering from the Massachusetts Institute of Technology (2017). Her work has been recognized with the IEEE Power &amp; Energy Society General Meeting Best Conference Paper Award (2022) and the American Control Conference Best Student Paper Award (2020). Amelia Musselman is an operations research engineer at Lawrence Livermore National Laboratory. She also has experience at Sandia National Laboratories, RAND Corporation, and Pacific Northwest National Laboratory. She holds a Ph.D. in Industrial Engineering and M.S. in Operations Research from Georgia Institute of Technology as well as a B.S. in Mathematics from Harvey Mudd College. Her expertise is in applications of optimization, including multi-objective, stochastic, and robust optimization, to critical infrastructure planning and protection. She has experience working on problems such as stochastic unit commitment, capacity expansion planning, power system and control network restoration, distributed optimization for electric vehicle charging, and other areas of power system planning and protection under uncertainty.<\/p> <p><strong>Tomas Valencia Zuluaga<\/strong>\u00a0is a postdoctoral researcher in the Center for Applied Scientific Computing at Lawrence Livermore National Laboratory (LLNL). His research lies at the intersection of optimization and power systems applications, including planning, operation and electricity markets. At LLNL, he has worked on developing high-performance-computing tools for optimal power grid expansion planning under uncertainty. He obtained his PhD in Industrial Engineering and Operations Research (2024) from UC Berkeley, where he received the IEOR Faculty Fellowship award (2023) and Outstanding Graduate Student Instructor Award (2022). He also holds a MS in Electrical Engineering (2018) and BS in Mechatronics Engineering (2014) from Universidad Nacional de Colombia in Bogot\u00e1, and has professional experience in the Oil &amp; Gas and Power sectors in Colombia.<\/p> <p><strong>Jean-Paul Watson (\"JP\")<\/strong>\u00a0is a Distinguished Member of Technical Staff in the Computational Engineering Division (CED) at Lawrence Livermore National Laboratory (LLNL), in Livermore California. At LLNL, he is the Associate Program Lead for Disaster Resilience in the Global Security Directorate. JP leads a diverse team of researchers focused on developing advanced analytics for critical infrastructure operations, planning, and resilience \u2013 emphasizing decision-making under uncertainty. He is co-inventor of the widely used Pyomo algebraic modeling language for mathematical optimization (www.pyomo.org), and has co-authored over 75 journal articles, 30 conference papers, and 3 books. JP has received the R&amp;D 100 award, the INFORMS Computing Society Prize, and was an INFORMS Edelman finalist.<\/p>\n<p><strong>Sunday, 4:15 PM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>\n<h3>Fraud Analytics as a Sequential Decision System: Integrating Machine Learning, Optimization, and Adversarial Learning<\/h3> <p><strong>Speaker: Tahir Ekin<\/strong><\/p>\n<ul><li><h4>Presentation and Speaker Bio<\/h4><p>Fraud analytics is fundamentally a decision problem under uncertainty,<br>involving trade-offs between detection performance, investigation costs, and operational constraints. While statistical and machine learning models are widely used to detect anomalous behavior and estimate fraud risk, they are often deployed without explicit consideration of downstream decisions, strategic adaptation, and system-level dynamics.<\/p> <p>This tutorial presents a unified framework that reframes fraud analytics as a sequential decision system involving adaptive and adversarial agents. We <br>review descriptive and predictive models, and embed them within a decision-theoretic framework that captures trade-offs between false positives, false negatives, and resource constraints. We further introduce adversarial risk analysis and related approaches to model strategic interactions between fraudsters and detection systems. By connecting prediction, optimization, and adversarial modeling, the tutorial highlights how decisions\u00a0influence both operational outcomes and future data through feedback effects.\u00a0This sequential perspective emphasizes the need for adaptive policies that account\u00a0for evolving fraud <br>behavior and changing system conditions. Examples from health\u00a0care and financial fraud illustrate how analytical models support real-world decision-making\u00a0and resource allocation in high-stakes environments. Intended for researchers\u00a0and practitioners across operations research, statistics, and data science, this tutorial\u00a0provides an accessible synthesis that requires no prior background. By the end, the\u00a0reader will understand how to frame fraud detection as a decision problem, evaluate the\u00a0limitations of purely predictive approaches, and reason about adversarial adaptation\u00a0within a unified sequential framework applicable across fraud domains.<\/p> <p><strong>Tahir Ekin<\/strong>\u00a0is Fields Chair in Business Analytics and Professor of Analytics at Texas State University, where he also serves as the founding Director of the Center for Analytics and Data Science (TXST CADS). His research focuses on probabilistic modeling, statistical learning, decision-making under uncertainty, adversarial machine learning, and fraud analytics, with applications in health care. He is the author of \"Statistics and Health Care Fraud: How to Save Billions\" (ASA\u2013CRC Press). His work has been supported by the National Science Foundation, Air Force Office of Scientific Research, and Texas Health and Human Services. He holds a Ph.D. in Decision Sciences from The George Washington University.<\/p><\/li><\/ul>\n<p>Fraud analytics is fundamentally a decision problem under uncertainty,<br>involving trade-offs between detection performance, investigation costs, and operational constraints. While statistical and machine learning models are widely used to detect anomalous behavior and estimate fraud risk, they are often deployed without explicit consideration of downstream decisions, strategic adaptation, and system-level dynamics.<\/p> <p>This tutorial presents a unified framework that reframes fraud analytics as a sequential decision system involving adaptive and adversarial agents. We <br>review descriptive and predictive models, and embed them within a decision-theoretic framework that captures trade-offs between false positives, false negatives, and resource constraints. We further introduce adversarial risk analysis and related approaches to model strategic interactions between fraudsters and detection systems. By connecting prediction, optimization, and adversarial modeling, the tutorial highlights how decisions\u00a0influence both operational outcomes and future data through feedback effects.\u00a0This sequential perspective emphasizes the need for adaptive policies that account\u00a0for evolving fraud <br>behavior and changing system conditions. Examples from health\u00a0care and financial fraud illustrate how analytical models support real-world decision-making\u00a0and resource allocation in high-stakes environments. Intended for researchers\u00a0and practitioners across operations research, statistics, and data science, this tutorial\u00a0provides an accessible synthesis that requires no prior background. By the end, the\u00a0reader will understand how to frame fraud detection as a decision problem, evaluate the\u00a0limitations of purely predictive approaches, and reason about adversarial adaptation\u00a0within a unified sequential framework applicable across fraud domains.<\/p> <p><strong>Tahir Ekin<\/strong>\u00a0is Fields Chair in Business Analytics and Professor of Analytics at Texas State University, where he also serves as the founding Director of the Center for Analytics and Data Science (TXST CADS). His research focuses on probabilistic modeling, statistical learning, decision-making under uncertainty, adversarial machine learning, and fraud analytics, with applications in health care. He is the author of \"Statistics and Health Care Fraud: How to Save Billions\" (ASA\u2013CRC Press). His work has been supported by the National Science Foundation, Air Force Office of Scientific Research, and Texas Health and Human Services. He holds a Ph.D. in Decision Sciences from The George Washington University.<\/p>\n<p><strong>Monday, 8:00 AM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>\n<h3>Multiagent Online Learning in Dynamic and <br>Uncertain Environments<\/h3> <p><strong>Speakers: Ceyhun Eksin, Jeff S. Shamma, Behrouz Touri<\/strong><\/p>\n<ul><li><h4>Presentation and Speaker Bios<\/h4>Multiagent online learning studies how multiple decision-making agents adapt their behavior over time in response to strategic interaction, uncertainty, and non-stationary environments created by other adaptive agents. Such settings arise naturally in large-scale engineered and socio-technical systems, including transportation networks, energy markets, financial systems, supply chains, and emerging agentic AI platforms, where agents may be any combination of humans, algorithms, or physical systems. \u00a0 This article presents a tutorial overview of learning in games and evolutionary game theory as a foundational framework for modeling and analyzing these interactions. We introduce core game-theoretic concepts and discuss how these outcomes may emerge under adaptive learning dynamics, along with selected impossibility results that capture obstacles to these outcomes. Representative discrete-time and continuous-time learning algorithms and their connections are reviewed alongside their convergence and long-run properties. The tutorial further presents learning for stochastic and Markov game settings, drawing connections to multiagent reinforcement learning and illustrating how strategic-form learning results can be leveraged in this generalized setting. \u00a0 <strong>Ceyhun Eksin <\/strong>is an Associate Professor with the Department of Industrial and Systems Engineering, Texas A&amp;M\u00a0University, College Station, TX, USA. He received the B.Sc. degree in Control Engineering from Istanbul Technical University, Istanbul, Turkey, in 2005, the M.S. degree in Industrial Engineering from Bogazi\u00e7i University, Istanbul, in 2008, the M.A. degree in statistics from the Wharton Statistics Department, and the Ph.D. degree in Electrical and Systems Engineering from University of Pennsylvania, in 2015. He was a Post-Doctoral Researcher jointly affiliated with the School of Biological Sciences and the School of Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, GA, USA. He was a recipient of the NSF CAREER Award in 2023. His research interests include networks, game theory, control theory, and distributed optimization. \u00a0 <strong>Jeff Shamma<\/strong> is the Department Head of Industrial and Enterprise Systems Engineering and Jerry S. Dobrovolny Chair at the University of Illinois Urbana-Champaign. He previously held faculty positions at the King Abdullah University of Science and Technology (KAUST) and at Georgia Tech as 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 past Distinguished Lecturer of the IEEE Control Systems Society, and a recipient of the IFAC High Impact Paper Award, AACC Donald P. Eckman Award, and NSF Young Investigator Award. Jeff has been a plenary\/semi-plenary speaker at NeurIPS, World Congress of the Game Theory Society, and IEEE Conference on Decision and Control. He was Editor-in-Chief of the IEEE Transactions on Control of Network Systems from 2020-2024. Jeff\u2019s research focuses on decision and control, game theory, and multi-agent systems. \u00a0 <strong>Behrouz Touri <\/strong>is an Associate Professor of the Industrial and Enterprise Systems Engineering at the University of Illinois at Urbana-Champaign (UIUC) and an affiliate faculty of the ECE Departments at UIUC and University of California, San Diego (UCSD). Prior to joining UIUC, he was an Associate Professor of ECE at the UCSD. He received his B.Sc. degree in Electrical Engineering from Isfahan University of Technology, Isfahan, Iran in 2006, his M.Sc. degree in Communications, Systems, Electronics from Jacobs University, Bremen, Germany in 2008, and his Ph.D. degree in Industrial Engineering from University of Illinois at Urbana-Champaign in 2011. His research interests include applied probability theory, distributed optimization, control and estimation, population dynamics, and game\u00a0theory.<\/li><\/ul>\nMultiagent online learning studies how multiple decision-making agents adapt their behavior over time in response to strategic interaction, uncertainty, and non-stationary environments created by other adaptive agents. Such settings arise naturally in large-scale engineered and socio-technical systems, including transportation networks, energy markets, financial systems, supply chains, and emerging agentic AI platforms, where agents may be any combination of humans, algorithms, or physical systems. \u00a0 This article presents a tutorial overview of learning in games and evolutionary game theory as a foundational framework for modeling and analyzing these interactions. We introduce core game-theoretic concepts and discuss how these outcomes may emerge under adaptive learning dynamics, along with selected impossibility results that capture obstacles to these outcomes. Representative discrete-time and continuous-time learning algorithms and their connections are reviewed alongside their convergence and long-run properties. The tutorial further presents learning for stochastic and Markov game settings, drawing connections to multiagent reinforcement learning and illustrating how strategic-form learning results can be leveraged in this generalized setting. \u00a0\n<strong>Ceyhun Eksin <\/strong>is an Associate Professor with the Department of Industrial and Systems Engineering, Texas A&amp;M\u00a0University, College Station, TX, USA. He received the B.Sc. degree in Control Engineering from Istanbul Technical University, Istanbul, Turkey, in 2005, the M.S. degree in Industrial Engineering from Bogazi\u00e7i University, Istanbul, in 2008, the M.A. degree in statistics from the Wharton Statistics Department, and the Ph.D. degree in Electrical and Systems Engineering from University of Pennsylvania, in 2015. He was a Post-Doctoral Researcher jointly affiliated with the School of Biological Sciences and the School of Electrical and Computer Engineering, Georgia Institute of Technology, Atlanta, GA, USA. He was a recipient of the NSF CAREER Award in 2023. His research interests include networks, game theory, control theory, and distributed optimization. \u00a0 <strong>Jeff Shamma<\/strong> is the Department Head of Industrial and Enterprise Systems Engineering and Jerry S. Dobrovolny Chair at the University of Illinois Urbana-Champaign. He previously held faculty positions at the King Abdullah University of Science and Technology (KAUST) and at Georgia Tech as 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 past Distinguished Lecturer of the IEEE Control Systems Society, and a recipient of the IFAC High Impact Paper Award, AACC Donald P. Eckman Award, and NSF Young Investigator Award. Jeff has been a plenary\/semi-plenary speaker at NeurIPS, World Congress of the Game Theory Society, and IEEE Conference on Decision and Control. He was Editor-in-Chief of the IEEE Transactions on Control of Network Systems from 2020-2024. Jeff\u2019s research focuses on decision and control, game theory, and multi-agent systems. \u00a0 <strong>Behrouz Touri <\/strong>is an Associate Professor of the Industrial and Enterprise Systems Engineering at the University of Illinois at Urbana-Champaign (UIUC) and an affiliate faculty of the ECE Departments at UIUC and University of California, San Diego (UCSD). Prior to joining UIUC, he was an Associate Professor of ECE at the UCSD. He received his B.Sc. degree in Electrical Engineering from Isfahan University of Technology, Isfahan, Iran in 2006, his M.Sc. degree in Communications, Systems, Electronics from Jacobs University, Bremen, Germany in 2008, and his Ph.D. degree in Industrial Engineering from University of Illinois at Urbana-Champaign in 2011. His research interests include applied probability theory, distributed optimization, control and estimation, population dynamics, and game\u00a0theory.\n<p><strong>Monday, 11:00 AM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>\n<h3>Multi-objective Combinatorial Optimization: Foundations, Theory, and Methods<\/h3> <p><strong>Speaker: Banu Lokman<\/strong><\/p>\n<ul><li><h4>Presentation and Speaker Bio<\/h4><p>This tutorial presents the foundations, theory and methods of multi-objective<br>integer programming, with a particular focus on multi-objective combinatorial<br>optimization (MOCO). Many real-world optimization problems are inherently<br>combinatorial in nature and involve multiple, often conflicting, objectives. MOCO has been widely used to model these complex decision problems across various domains such as transportation, logistics, finance, energy, and healthcare. While the feasible set is typically finite in MOCO problems, these solutions are not always explicitly available to the decision-makers. Adding to this complexity, there is rarely a single solution that optimizes all objectives simultaneously in multi-objective optimization problems.<\/p> <p>This tutorial provides a structured framework for students, researchers, and practitioners to approach and solve these complex decision problems. We first define key terminology, describe the main characteristics of MOCO problems and discuss their scalarization. We then explore the advanced methods and exact algorithms to generate all nondominated points, for which an improvement in one objective cannot be made without sacrificing performance in another. Since these algorithms become\u00a0intractable in real-world problem settings with the increase in the number of nondominated\u00a0points, we also discuss the methods that generate a representative set of\u00a0solutions with a prespecified level of quality or find preferred solutions. The tutorial\u00a0provides a broad and accessible overview of existing methods, supported by illustrative\u00a0examples, discussions, figures, and comprehensive references that clarify their\u00a0main ideas, strengths, and limitations.<\/p> <p><strong>Banu Lokman<\/strong> is a Professor of Operational Research in the School of Organisations, Systems and People at the University of Portsmouth, United Kingdom. She completed her PhD in Industrial Engineering at Middle East Technical University (METU) in Turkey and held academic and research positions at METU and the Aalto University School of Business in Finland. She currently serves as the President of the INFORMS Section on Multiple Criteria Decision Making (MCDM), having previously held board positions in the International Society on MCDM and the INFORMS MCDM Section. She is an Associate Editor for Omega and the IMA Journal of Management Mathematics and is an editorial board member of the Journal of Multi-Criteria Decision Analysis. She is also a member of the Research Committee of the UK Operational Research Society and leads UK-based MCDM courses at NATCOR. In 2022, she received the Bernard Roy Award from the Association of European Operational Research Societies (EURO) Working Group on Multiple Criteria Decision Aiding for her contributions to the field. Her research interests include MCDM, combinatorial optimization, multi-objective integer and mixed-integer programming, clustering, and applications in energy, sustainability, and healthcare. Her work has been published in several peer-reviewed journals, including Management Science and European Journal of Operational Research.<\/p><\/li><\/ul>\n<p>This tutorial presents the foundations, theory and methods of multi-objective<br>integer programming, with a particular focus on multi-objective combinatorial<br>optimization (MOCO). Many real-world optimization problems are inherently<br>combinatorial in nature and involve multiple, often conflicting, objectives. MOCO has been widely used to model these complex decision problems across various domains such as transportation, logistics, finance, energy, and healthcare. While the feasible set is typically finite in MOCO problems, these solutions are not always explicitly available to the decision-makers. Adding to this complexity, there is rarely a single solution that optimizes all objectives simultaneously in multi-objective optimization problems.<\/p> <p>This tutorial provides a structured framework for students, researchers, and practitioners to approach and solve these complex decision problems. We first define key terminology, describe the main characteristics of MOCO problems and discuss their scalarization. We then explore the advanced methods and exact algorithms to generate all nondominated points, for which an improvement in one objective cannot be made without sacrificing performance in another. Since these algorithms become\u00a0intractable in real-world problem settings with the increase in the number of nondominated\u00a0points, we also discuss the methods that generate a representative set of\u00a0solutions with a prespecified level of quality or find preferred solutions. The tutorial\u00a0provides a broad and accessible overview of existing methods, supported by illustrative\u00a0examples, discussions, figures, and comprehensive references that clarify their\u00a0main ideas, strengths, and limitations.<\/p> <p><strong>Banu Lokman<\/strong> is a Professor of Operational Research in the School of Organisations, Systems and People at the University of Portsmouth, United Kingdom. She completed her PhD in Industrial Engineering at Middle East Technical University (METU) in Turkey and held academic and research positions at METU and the Aalto University School of Business in Finland. She currently serves as the President of the INFORMS Section on Multiple Criteria Decision Making (MCDM), having previously held board positions in the International Society on MCDM and the INFORMS MCDM Section. She is an Associate Editor for Omega and the IMA Journal of Management Mathematics and is an editorial board member of the Journal of Multi-Criteria Decision Analysis. She is also a member of the Research Committee of the UK Operational Research Society and leads UK-based MCDM courses at NATCOR. In 2022, she received the Bernard Roy Award from the Association of European Operational Research Societies (EURO) Working Group on Multiple Criteria Decision Aiding for her contributions to the field. Her research interests include MCDM, combinatorial optimization, multi-objective integer and mixed-integer programming, clustering, and applications in energy, sustainability, and healthcare. Her work has been published in several peer-reviewed journals, including Management Science and European Journal of Operational Research.<\/p>\n<p><strong>Monday, 1:15 PM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>\n<h3>Self-Adapting Approximations of Markov Decision Processes: A Guided Tour<\/h3> <p><strong>Speakers: Andre Augusto Cire, Selvaprabu Nadarajah, Parshan Pakiman, Negar Soheili<\/strong><\/p>\n<ul><li><h4>Presentation and Speaker Bios<\/h4><p>Sequential decision making under uncertainty arises across business, <br>engineering, and science. Markov Decision Processes (MDPs) offer a rich modeling language, but translating models into implementable policies requires confronting challenging state and action spaces and non-convex optimization landscapes. Approximations are essential, and making them accessible and performant without algorithmic expertise allows users to focus on formulating MDPs that capture real-world features rather than\u00a0simplifying their models to facilitate solution. Obtaining a policy involves a three-stage COR cycle: Constructing a parametrized model that approximates the MDP, Optimizing for the parameters of this model, and Refining the model by learning from the\u00a0solution. These stages involve design choices that burden the user, leading them to simplify or skip stages.<\/p> <p>This tutorial provides an overview of methods that reduce the user burden using COR as its guide. Construction ranges from lightweight extensible models to automated structural reformulations. Optimization leverages mathematical programming and first-order methods that have matured in theory and off-the-shelf software. Refinement performs feature expansion and approximation tightening, feeding back into the construct stage. When a combination of techniques across COR stages results in both accessibility and performance, we refer to it as a self-adapting approximation. Such an approximation adapts to problem structure, instance data, or solution progress. We guide readers through these ideas across weakly coupled MDPs and general continuous state-action MDPs, providing generalizable foundations and a code base to get started on a simple example. The tutorial concludes by outlining the potential for foundation models and quantum computing to further impact this area.<\/p> <p><strong>Andre A. Cire<\/strong> is an Associate Professor of Operations Management and Analytics at the University of Toronto, cross-appointed between the Rotman School of Management and the Department of Management at the Scarborough campus. His research bridges methodological development and practical applications in optimization, with a focus on sequential decision-making, mathematical programming, dynamic programming, network models, and applications in scheduling, healthcare, and supply chains. Andre earned his PhD in Operations Research from Carnegie Mellon University, and his contributions have been recognized with awards such as the INFORMS Harvey J. Greenberg Award (2024), an honorable mention for the INFORMS Computing Society Prize (2023), and research and teaching awards at the University of Toronto. He has also held senior editorial and service roles for multiple journals and conferences including INFORMS Journal on Computing, AAAI, and NeurIPS.<\/p> <p><strong>Selvaprabu (Selva) Nadarajah<\/strong> is an Associate Professor of Information and Decision Sciences and Bielinski Family Endowed Scholar at the College of Business Administration, University of Illinois Chicago (UIC). He is also a UIC Global Scholar and previously served as the Decision Intelligence R&amp;D Lead at the Discovery Partners Institute, the innovation hub of the University of Illinois System. He is an Associate Editor for Decision Sciences, Operations Research, and Production and Operations Management. His research develops self-adapting approximation methods for large-scale Markov decision processes. He applies these methods to energy investment and operations problems, including real options valuation of commodity and energy assets, renewable energy procurement, and capacity expansion. This work has been recognized with awards from INFORMS, the Commodity and Energy Markets Association, and NeurIPS. It has been funded by the Alfred P. Sloan Foundation, Argonne National Laboratory, and industry. He has received the UIC College of Business Teaching Excellence Award and the EnergyTech University Prize Faculty Explorer Award from the U.S. Department of Energy for educational initiatives. He holds a PhD in Operations Research from the Tepper School of Business at Carnegie Mellon University.<\/p> <p><strong>Parshan Pakiman<\/strong> is an Assistant Professor in the Department of Operations Management and Strategy at the University at Buffalo School of Management. His research focuses on approximate dynamic programming and data-driven optimization methods for sequential decision making, with an emphasis on large-scale Markov decision processes and structured relaxations. His work develops algorithms with strong theoretical guarantees that minimize the need for human intervention during deployment, including model selection and parameter tuning. These methods are designed to be scalable, interpretable, and accessible to practitioners, with applications in healthcare operations and revenue management. Prior to joining the University at Buffalo, he was a Principal Researcher at the Tolan Center for Healthcare at the University of Chicago Booth School of Business. He received his PhD in Information and Decision Sciences from the University of Illinois Chicago and his BSc in Applied Mathematics from the University of Tehran.<\/p> <p><strong>Negar Soheili<\/strong> is an Associate Professor in the Department of Information and Decision Sciences at the University of Illinois Chicago. She received her PhD in Operations Research from Carnegie Mellon University. Her research develops scalable optimization algorithms for machine learning and sequential decision-making under uncertainty. In particular, she designs efficient first-order methods for large-scale constrained optimization and the approximation of Markov decision processes, with applications in healthcare, supply chains, and business analytics. She launched the PhD program in Operations Research and Business Analytics at UIC. She serves as an Associate Editor for the INFORMS Journal on Computing and for Optimization Letters.<\/p><\/li><\/ul>\n<p>Sequential decision making under uncertainty arises across business, <br>engineering, and science. Markov Decision Processes (MDPs) offer a rich modeling language, but translating models into implementable policies requires confronting challenging state and action spaces and non-convex optimization landscapes. Approximations are essential, and making them accessible and performant without algorithmic expertise allows users to focus on formulating MDPs that capture real-world features rather than\u00a0simplifying their models to facilitate solution. Obtaining a policy involves a three-stage COR cycle: Constructing a parametrized model that approximates the MDP, Optimizing for the parameters of this model, and Refining the model by learning from the\u00a0solution. These stages involve design choices that burden the user, leading them to simplify or skip stages.<\/p> <p>This tutorial provides an overview of methods that reduce the user burden using COR as its guide. Construction ranges from lightweight extensible models to automated structural reformulations. Optimization leverages mathematical programming and first-order methods that have matured in theory and off-the-shelf software. Refinement performs feature expansion and approximation tightening, feeding back into the construct stage. When a combination of techniques across COR stages results in both accessibility and performance, we refer to it as a self-adapting approximation. Such an approximation adapts to problem structure, instance data, or solution progress. We guide readers through these ideas across weakly coupled MDPs and general continuous state-action MDPs, providing generalizable foundations and a code base to get started on a simple example. The tutorial concludes by outlining the potential for foundation models and quantum computing to further impact this area.<\/p> <p><strong>Andre A. Cire<\/strong> is an Associate Professor of Operations Management and Analytics at the University of Toronto, cross-appointed between the Rotman School of Management and the Department of Management at the Scarborough campus. His research bridges methodological development and practical applications in optimization, with a focus on sequential decision-making, mathematical programming, dynamic programming, network models, and applications in scheduling, healthcare, and supply chains. Andre earned his PhD in Operations Research from Carnegie Mellon University, and his contributions have been recognized with awards such as the INFORMS Harvey J. Greenberg Award (2024), an honorable mention for the INFORMS Computing Society Prize (2023), and research and teaching awards at the University of Toronto. He has also held senior editorial and service roles for multiple journals and conferences including INFORMS Journal on Computing, AAAI, and NeurIPS.<\/p> <p><strong>Selvaprabu (Selva) Nadarajah<\/strong> is an Associate Professor of Information and Decision Sciences and Bielinski Family Endowed Scholar at the College of Business Administration, University of Illinois Chicago (UIC). He is also a UIC Global Scholar and previously served as the Decision Intelligence R&amp;D Lead at the Discovery Partners Institute, the innovation hub of the University of Illinois System. He is an Associate Editor for Decision Sciences, Operations Research, and Production and Operations Management. His research develops self-adapting approximation methods for large-scale Markov decision processes. He applies these methods to energy investment and operations problems, including real options valuation of commodity and energy assets, renewable energy procurement, and capacity expansion. This work has been recognized with awards from INFORMS, the Commodity and Energy Markets Association, and NeurIPS. It has been funded by the Alfred P. Sloan Foundation, Argonne National Laboratory, and industry. He has received the UIC College of Business Teaching Excellence Award and the EnergyTech University Prize Faculty Explorer Award from the U.S. Department of Energy for educational initiatives. He holds a PhD in Operations Research from the Tepper School of Business at Carnegie Mellon University.<\/p> <p><strong>Parshan Pakiman<\/strong> is an Assistant Professor in the Department of Operations Management and Strategy at the University at Buffalo School of Management. His research focuses on approximate dynamic programming and data-driven optimization methods for sequential decision making, with an emphasis on large-scale Markov decision processes and structured relaxations. His work develops algorithms with strong theoretical guarantees that minimize the need for human intervention during deployment, including model selection and parameter tuning. These methods are designed to be scalable, interpretable, and accessible to practitioners, with applications in healthcare operations and revenue management. Prior to joining the University at Buffalo, he was a Principal Researcher at the Tolan Center for Healthcare at the University of Chicago Booth School of Business. He received his PhD in Information and Decision Sciences from the University of Illinois Chicago and his BSc in Applied Mathematics from the University of Tehran.<\/p> <p><strong>Negar Soheili<\/strong> is an Associate Professor in the Department of Information and Decision Sciences at the University of Illinois Chicago. She received her PhD in Operations Research from Carnegie Mellon University. Her research develops scalable optimization algorithms for machine learning and sequential decision-making under uncertainty. In particular, she designs efficient first-order methods for large-scale constrained optimization and the approximation of Markov decision processes, with applications in healthcare, supply chains, and business analytics. She launched the PhD program in Operations Research and Business Analytics at UIC. She serves as an Associate Editor for the INFORMS Journal on Computing and for Optimization Letters.<\/p>\n<p><strong>Monday, 2:45 PM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>\n<h3>Decision-Focused Learning: When and Why Traditional Prediction Models Fail<\/h3> <p><strong>Speaker: Mo Liu<\/strong><\/p>\n<ul><li><h4>Presentation and Speaker Bio<\/h4><p>Plugging predictions of unknown parameters into downstream optimization<br>problems, often referred to as the \u201cpredict-then-optimize\u201d paradigm, has long<br>been a standard approach in decision-making under uncertainty. However, improved\u00a0predictive accuracy does not, in general, translate into improved decision quality.\u00a0This disconnect has motivated growing interest in decision-focused learning (DFL)\u00a0within the operations research community.<\/p> <p>This tutorial reviews recent developments\u00a0in DFL and highlights key methodological insights, with a particular focus on stochastic\u00a0linear programming as the downstream decision-making problem. We discuss why\u00a0several widely used tools in traditional statistical learning are not directly suited to\u00a0decision-focused settings and must be rethought, including (i) data collection strategies\u00a0driven purely by predictive uncertainty and (ii) distributional distance measures\u00a0such as the Wasserstein distance. We summarize properties of DFL that distinguish it\u00a0from conventional predictive modeling and provide insights into the development of\u00a0new decision-focused tools.<\/p> <p><strong>Mo Liu<\/strong> is an assistant professor in the Department of Statistics and Operations Research at the University of North Carolina at Chapel Hill. He received his PhD in Industrial Engineering and Operations Research from the University of California, Berkeley. His research centers on decision-focused learning, a methodology that designs and trains prediction models while accounting for downstream optimization problems. These downstream problems arise in real-world applications such as revenue and inventory management.<\/p><\/li><\/ul>\n<p>Plugging predictions of unknown parameters into downstream optimization<br>problems, often referred to as the \u201cpredict-then-optimize\u201d paradigm, has long<br>been a standard approach in decision-making under uncertainty. However, improved\u00a0predictive accuracy does not, in general, translate into improved decision quality.\u00a0This disconnect has motivated growing interest in decision-focused learning (DFL)\u00a0within the operations research community.<\/p> <p>This tutorial reviews recent developments\u00a0in DFL and highlights key methodological insights, with a particular focus on stochastic\u00a0linear programming as the downstream decision-making problem. We discuss why\u00a0several widely used tools in traditional statistical learning are not directly suited to\u00a0decision-focused settings and must be rethought, including (i) data collection strategies\u00a0driven purely by predictive uncertainty and (ii) distributional distance measures\u00a0such as the Wasserstein distance. We summarize properties of DFL that distinguish it\u00a0from conventional predictive modeling and provide insights into the development of\u00a0new decision-focused tools.<\/p> <p><strong>Mo Liu<\/strong> is an assistant professor in the Department of Statistics and Operations Research at the University of North Carolina at Chapel Hill. He received his PhD in Industrial Engineering and Operations Research from the University of California, Berkeley. His research centers on decision-focused learning, a methodology that designs and trains prediction models while accounting for downstream optimization problems. These downstream problems arise in real-world applications such as revenue and inventory management.<\/p>\n<p><strong>Monday, 4:15 PM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>\n<h3>Generative Models for Decision-Making under Distributional Shift<\/h3> <p><strong>Speakers: Xiuyuan Cheng, Yunqin Zhu, Yao Xie<\/strong><\/p>\n<ul><li><h4>Presentation and Speaker Bios<\/h4>Many data-driven decision problems are formulated using a nominal distribution<br>estimated from historical data, while performance is ultimately determined by<br>a deployment distribution that may be shifted, context-dependent, partially observed,\u00a0or stress-induced. This tutorial presents modern generative models, particularly flow and\u00a0score-based methods, as mathematical tools for constructing decision-relevant\u00a0distributions. From an operations research perspective, their primary value lies not\u00a0in unconstrained sample synthesis but in representing and transforming distributions\u00a0through transport maps, velocity fields, score fields, and guided stochastic dynamics.\u00a0We present a unified framework based on pushforward maps, continuity, Fokker\u2013\u00a0Planck equations, Wasserstein geometry, and optimization in probability space. \u00a0 Within this framework, generative models can be used to learn nominal uncertainty, construct stressed or least-favorable distributions for robustness, and produce conditional or\u00a0posterior distributions under side information and partial observation. We also highlight\u00a0representative theoretical guarantees, including forward\u2013reverse convergence for\u00a0iterative flow models, first-order minimax analysis in transport-map space, and error transfer\u00a0bounds for posterior sampling with generative priors. The tutorial provides\u00a0a principled introduction to using generative models for scenario generation, robust\u00a0decision-making, uncertainty quantification, and related problems under distributional\u00a0shift. \u00a0 <p>Xiuyuan Cheng is a professor of mathematics at Duke University. She received her Ph.D. degree from Princeton University in 2013. Before joining Duke, she was a postdoctoral researcher at \u00c9cole Normale Sup\u00e9rieure in Paris from 2013 to 2015 and a Gibbs Assistant Professor at Yale University from 2015 to 2017.\u00a0Her research interests include theoretical and computational techniques for high-dimensional data\u00a0analysis, signal processing, and machine learning. She is a recipient of a Sloan Fellowship and an\u00a0NSF CAREER Award.<\/p> <p><strong>Yunqin \u201cVinci\u201d Zhu<\/strong> is a Ph.D. student in the H. Milton Stewart School of Industrial and Systems Engineering at the Georgia Institute of Technology. He received his B.Eng. degree in artificial intelligence from the University of Science and Technology (USTC) of China. His research interests include machine learning, generative modeling, and optimization. He is a recipient of the Guo Moruo Scholarship from USTC.<\/p> <p><strong>Yao Xie<\/strong>\u00a0 is the Coca-Cola\u00a0Foundation Chair and Professor in the H. Milton Stewart School of Industrial and Systems Engineering at the Georgia Institute of Technology, where she is also Associate Director of the Machine Learning Center. She received her Ph.D. degree in electrical engineering, with a minor in mathematics, from Stanford University and was previously a Research Scientist at Duke University.\u00a0Her research lies at the intersection of statistics, machine learning, and optimization, with a focus\u00a0on developing statistically powerful and computationally efficient methods for high-dimensional,\u00a0sequential, and spatio-temporal data. She is a Member of Cohort 2026 of the National Academies\u2019\u00a0New Voices in Sciences, Engineering, and Medicine program and the IEEE Information Theory\u00a0Society Distinguished Lecturer for 2026\u20132027. Her honors include the NSF CAREER Award,\u00a0INFORMS Wagner Prize Finalist, INFORMS Gaver Early Career Award, and C.W.S. Woodroofe\u00a0Award. She serves as an associate editor for several journals, including IEEE Transactions on\u00a0Information Theory, Journal of the American Statistical Association\u2014Theory and Methods, Operations\u00a0Research, Annals of Applied Statistics, Sequential Analysis, and INFORMS Journal on Data\u00a0Science.<\/p><\/li><\/ul>\nMany data-driven decision problems are formulated using a nominal distribution<br>estimated from historical data, while performance is ultimately determined by<br>a deployment distribution that may be shifted, context-dependent, partially observed,\u00a0or stress-induced. This tutorial presents modern generative models, particularly flow and\u00a0score-based methods, as mathematical tools for constructing decision-relevant\u00a0distributions. From an operations research perspective, their primary value lies not\u00a0in unconstrained sample synthesis but in representing and transforming distributions\u00a0through transport maps, velocity fields, score fields, and guided stochastic dynamics.\u00a0We present a unified framework based on pushforward maps, continuity, Fokker\u2013\u00a0Planck equations, Wasserstein geometry, and optimization in probability space. \u00a0 Within this framework, generative models can be used to learn nominal uncertainty, construct stressed or least-favorable distributions for robustness, and produce conditional or\u00a0posterior distributions under side information and partial observation. We also highlight\u00a0representative theoretical guarantees, including forward\u2013reverse convergence for\u00a0iterative flow models, first-order minimax analysis in transport-map space, and error transfer\u00a0bounds for posterior sampling with generative priors. The tutorial provides\u00a0a principled introduction to using generative models for scenario generation, robust\u00a0decision-making, uncertainty quantification, and related problems under distributional\u00a0shift. \u00a0\n<p>Xiuyuan Cheng is a professor of mathematics at Duke University. She received her Ph.D. degree from Princeton University in 2013. Before joining Duke, she was a postdoctoral researcher at \u00c9cole Normale Sup\u00e9rieure in Paris from 2013 to 2015 and a Gibbs Assistant Professor at Yale University from 2015 to 2017.\u00a0Her research interests include theoretical and computational techniques for high-dimensional data\u00a0analysis, signal processing, and machine learning. She is a recipient of a Sloan Fellowship and an\u00a0NSF CAREER Award.<\/p> <p><strong>Yunqin \u201cVinci\u201d Zhu<\/strong> is a Ph.D. student in the H. Milton Stewart School of Industrial and Systems Engineering at the Georgia Institute of Technology. He received his B.Eng. degree in artificial intelligence from the University of Science and Technology (USTC) of China. His research interests include machine learning, generative modeling, and optimization. He is a recipient of the Guo Moruo Scholarship from USTC.<\/p> <p><strong>Yao Xie<\/strong>\u00a0 is the Coca-Cola\u00a0Foundation Chair and Professor in the H. Milton Stewart School of Industrial and Systems Engineering at the Georgia Institute of Technology, where she is also Associate Director of the Machine Learning Center. She received her Ph.D. degree in electrical engineering, with a minor in mathematics, from Stanford University and was previously a Research Scientist at Duke University.\u00a0Her research lies at the intersection of statistics, machine learning, and optimization, with a focus\u00a0on developing statistically powerful and computationally efficient methods for high-dimensional,\u00a0sequential, and spatio-temporal data. She is a Member of Cohort 2026 of the National Academies\u2019\u00a0New Voices in Sciences, Engineering, and Medicine program and the IEEE Information Theory\u00a0Society Distinguished Lecturer for 2026\u20132027. Her honors include the NSF CAREER Award,\u00a0INFORMS Wagner Prize Finalist, INFORMS Gaver Early Career Award, and C.W.S. Woodroofe\u00a0Award. She serves as an associate editor for several journals, including IEEE Transactions on\u00a0Information Theory, Journal of the American Statistical Association\u2014Theory and Methods, Operations\u00a0Research, Annals of Applied Statistics, Sequential Analysis, and INFORMS Journal on Data\u00a0Science.<\/p>\n<p><strong>Tuesday, 11:00 AM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>\n<h3>GPU-Accelerated Decision Optimization<\/h3> <p><strong>Speakers: Nicolas Blin, Burcin Bozkaya, Akif \u00c7\u00f6rd\u00fck, Chris Maes<\/strong><\/p>\n<ul><li><h4>Presentation and Speaker Bios<\/h4><p>For decades, algorithms for solving decision optimization problems have been designed and implemented for CPUs.\u00a0Recent advances in GPU hardware, driven by machine learning and AI, together with the development of GPU-accelerated scientific computing\u00a0kernels, have made GPUs a practical platform for solving optimization problems. This shift is most visible in linear\u00a0programming, where first-order methods exploit GPU parallelism and high-bandwidth memory to solve problems with millions\u00a0of variables and constraints. GPU algorithms for mixed-integer programming are still emerging, while GPU-based vehicle-routing<br>solvers are more mature and already support large-scale routing applications. At the same time, generative AI and agentic AI\u00a0are changing how optimization models are formulated, solved, and embedded in decision workflows.<\/p> <p>This tutorial introduces\u00a0GPU-accelerated decision optimization for researchers and practitioners familiar with CPU-based modeling and solvers. We review<br>computational patterns that make optimization algorithms amenable to GPU acceleration, summarize the growing ecosystem of opensource\u00a0and commercial GPU solvers, and provide benchmarks that quantify current performance. We also present cuOpt, NVIDIA\u2019s\u00a0open-source library for GPU-accelerated decision optimization. Our goal is to help the optimization community understand where\u00a0GPUs are useful today, where important limitations remain, and where future research and development are needed.<\/p> <p><strong>Nicolas Blin<\/strong>\u00a0is a senior developer technology engineer at NVIDIA. Nicolas holds an M.Sc. in software engineering. He has a background in image processing and GPU-accelerated algorithms. His interests include massively parallel algorithms, performance optimization, and combinatorial and linear optimization.<\/p> <p><strong>Burcin Bozkaya\u00a0<\/strong>is a senior developer relations manager at NVIDIA. He holds a BS and MS in Industrial Engineering and PhD in Management Science, specializing in combinatorial optimization problems and heuristic optimization as applied in transportation and logistics, and supply chain planning. As the developer relations lead for decision optimization, Burcin engages with the OR developer ecosystem to evangelize and support the open-source community, ISV partners and key business planners for developing accelerated optimization solvers.<\/p> <p><strong>Akif Coerduek<\/strong>\u00a0is a senior developer technology engineer at NVIDIA. Akif holds a bachelor\u2019s degree in computer engineering and a master\u2019s degree in software engineering. He has a background in massively parallel pricing engines and GPU-accelerated combinatorial optimization. His interests include parallel algorithms, performance optimization, and combinatorial optimization.<\/p> <p><strong>Christopher Maes<\/strong>\u00a0is a principal engineer on the NVIDIA cuOpt team. He received his PhD in Computational and Mathematical Engineering from Stanford University in 2010. His background and research interests are in linear, quadratic, and mixed-integer programming, as well as trajectory optimization, and machine learning.<\/p><\/li><\/ul>\n<p>For decades, algorithms for solving decision optimization problems have been designed and implemented for CPUs.\u00a0Recent advances in GPU hardware, driven by machine learning and AI, together with the development of GPU-accelerated scientific computing\u00a0kernels, have made GPUs a practical platform for solving optimization problems. This shift is most visible in linear\u00a0programming, where first-order methods exploit GPU parallelism and high-bandwidth memory to solve problems with millions\u00a0of variables and constraints. GPU algorithms for mixed-integer programming are still emerging, while GPU-based vehicle-routing<br>solvers are more mature and already support large-scale routing applications. At the same time, generative AI and agentic AI\u00a0are changing how optimization models are formulated, solved, and embedded in decision workflows.<\/p> <p>This tutorial introduces\u00a0GPU-accelerated decision optimization for researchers and practitioners familiar with CPU-based modeling and solvers. We review<br>computational patterns that make optimization algorithms amenable to GPU acceleration, summarize the growing ecosystem of opensource\u00a0and commercial GPU solvers, and provide benchmarks that quantify current performance. We also present cuOpt, NVIDIA\u2019s\u00a0open-source library for GPU-accelerated decision optimization. Our goal is to help the optimization community understand where\u00a0GPUs are useful today, where important limitations remain, and where future research and development are needed.<\/p> <p><strong>Nicolas Blin<\/strong>\u00a0is a senior developer technology engineer at NVIDIA. Nicolas holds an M.Sc. in software engineering. He has a background in image processing and GPU-accelerated algorithms. His interests include massively parallel algorithms, performance optimization, and combinatorial and linear optimization.<\/p> <p><strong>Burcin Bozkaya\u00a0<\/strong>is a senior developer relations manager at NVIDIA. He holds a BS and MS in Industrial Engineering and PhD in Management Science, specializing in combinatorial optimization problems and heuristic optimization as applied in transportation and logistics, and supply chain planning. As the developer relations lead for decision optimization, Burcin engages with the OR developer ecosystem to evangelize and support the open-source community, ISV partners and key business planners for developing accelerated optimization solvers.<\/p> <p><strong>Akif Coerduek<\/strong>\u00a0is a senior developer technology engineer at NVIDIA. Akif holds a bachelor\u2019s degree in computer engineering and a master\u2019s degree in software engineering. He has a background in massively parallel pricing engines and GPU-accelerated combinatorial optimization. His interests include parallel algorithms, performance optimization, and combinatorial optimization.<\/p> <p><strong>Christopher Maes<\/strong>\u00a0is a principal engineer on the NVIDIA cuOpt team. He received his PhD in Computational and Mathematical Engineering from Stanford University in 2010. His background and research interests are in linear, quadratic, and mixed-integer programming, as well as trajectory optimization, and machine learning.<\/p>\n<p><strong>Tuesday, 1:15 PM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>\n<h3>A Modern Treatment of the Primal\u2013Dual Framework for Online Resource Allocation<\/h3> <p><strong>Speakers: Rad Niazadeh, Rajan Udwani<\/strong><\/p>\n<ul><li><h4>Presentation and Speaker Bios<\/h4><p>Linear-programming (LP)-based primal\u2013dual methods are fundamental for designing and analyzing algorithms in adversarial (prior-free) online resource allocation. This chapter provides a tutorial on two modern primal-dual frameworks, emphasizing recent developments and contemporary models in operations research.<\/p> <p>Part I develops an LP-based convex-programming framework where solving a regularized convex program at each arrival captures the tradeoff between greediness and hedging, yielding a dual certificate via Karush\u2013Kuhn\u2013Tucker\u00a0(KKT) conditions. Because standard LP relaxations can be weak or intractable\u00a0for stochastic outcomes, Part II introduces a complementary LP-free framework that provides a universal certificate system for evaluating competitive ratios\u00a0under such uncertainty. Covering a wide array of models\u2014including online vertex-weighted bipartite matching, edge-weighted online matching with free disposal, online matching with stochastic rewards, reusable resources, two-sided assortment optimization, configuration allocation (whole-page optimization), AdWords, and costly cancellations\u2014the tutorial equips readers with versatile <br>proof templates to analyze existing algorithms and develop new solutions for emerging applications.<\/p> <strong>Rad Niazadeh<\/strong> is an Associate Professor of Operations Management at the University of Chicago Booth School of Business. He is also part of the faculty at Toyota Technological Institute of Chicago (TTIC) by a courtesy appointment, and a faculty advisor at Lyft Inc. (Fulfillment &amp; Matching Science). Prior to joining Chicago Booth, he was a visiting researcher at Google Research NYC and a Motwani postdoctoral fellow at Stanford University, Computer Science Department. He obtained his PhD in Computer Science (minored in Applied Mathematics) from Cornell University. His research spans online algorithms, online learning, and mechanism design, with applications in theory and practice of online platforms and non-profit operations. His work has received multiple recognitions including the INFORMS Auctions and Market Design Rothkopf Junior Researcher Paper Award (2021, 2024: first place; 2023: second place), the INFORMS Revenue Management and Pricing Dissertation Award (honorable mention), and best student paper awards at INFORMS George Nicholson Competition (2025: honorable mention), INFORMS MSOM (2024), INFORMS Service Science (2025: finalist), INFORMS Data Mining, and IJCAI conference. \u00a0 <strong>Rajan Udwani<\/strong> is an Assistant Professor of Industrial Engineering and Operations Research at the University of California, Berkeley. His research focuses on algorithms for optimization under uncertainty, with a particular emphasis on revenue management and pricing. Before joining UC Berkeley, he was a postdoctoral researcher at Columbia University. He holds a Ph.D. in Operations Research from MIT and a B.Tech. in Electrical Engineering from IIT Bombay. <br>His work has been recognized with INFORMS junior faculty and student paper awards, and is supported by an NSF CAREER Award and the Google Research Scholar Program.<\/li><\/ul>\n<p>Linear-programming (LP)-based primal\u2013dual methods are fundamental for designing and analyzing algorithms in adversarial (prior-free) online resource allocation. This chapter provides a tutorial on two modern primal-dual frameworks, emphasizing recent developments and contemporary models in operations research.<\/p> <p>Part I develops an LP-based convex-programming framework where solving a regularized convex program at each arrival captures the tradeoff between greediness and hedging, yielding a dual certificate via Karush\u2013Kuhn\u2013Tucker\u00a0(KKT) conditions. Because standard LP relaxations can be weak or intractable\u00a0for stochastic outcomes, Part II introduces a complementary LP-free framework that provides a universal certificate system for evaluating competitive ratios\u00a0under such uncertainty. Covering a wide array of models\u2014including online vertex-weighted bipartite matching, edge-weighted online matching with free disposal, online matching with stochastic rewards, reusable resources, two-sided assortment optimization, configuration allocation (whole-page optimization), AdWords, and costly cancellations\u2014the tutorial equips readers with versatile <br>proof templates to analyze existing algorithms and develop new solutions for emerging applications.<\/p> <strong>Rad Niazadeh<\/strong> is an Associate Professor of Operations Management at the University of Chicago Booth School of Business. He is also part of the faculty at Toyota Technological Institute of Chicago (TTIC) by a courtesy appointment, and a faculty advisor at Lyft Inc. (Fulfillment &amp; Matching Science). Prior to joining Chicago Booth, he was a visiting researcher at Google Research NYC and a Motwani postdoctoral fellow at Stanford University, Computer Science Department. He obtained his PhD in Computer Science (minored in Applied Mathematics) from Cornell University. His research spans online algorithms, online learning, and mechanism design, with applications in theory and practice of online platforms and non-profit operations. His work has received multiple recognitions including the INFORMS Auctions and Market Design Rothkopf Junior Researcher Paper Award (2021, 2024: first place; 2023: second place), the INFORMS Revenue Management and Pricing Dissertation Award (honorable mention), and best student paper awards at INFORMS George Nicholson Competition (2025: honorable mention), INFORMS MSOM (2024), INFORMS Service Science (2025: finalist), INFORMS Data Mining, and IJCAI conference. \u00a0 <strong>Rajan Udwani<\/strong> is an Assistant Professor of Industrial Engineering and Operations Research at the University of California, Berkeley. His research focuses on algorithms for optimization under uncertainty, with a particular emphasis on revenue management and pricing. Before joining UC Berkeley, he was a postdoctoral researcher at Columbia University. He holds a Ph.D. in Operations Research from MIT and a B.Tech. in Electrical Engineering from IIT Bombay. <br>His work has been recognized with INFORMS junior faculty and student paper awards, and is supported by an NSF CAREER Award and the Google Research Scholar Program.\n<p><strong>Tuesday, 2:45 PM<\/strong><br><strong>Moscone South-156<\/strong><br><strong>(Upper Mezz)<\/strong><\/p>\n<h3>Foundations of Reinforcement Learning and Control: Connections and New Perspective<\/h3> <p><strong>Speakers: Claire Vernade, Onno Eberhard Max, Martha White, Florian D\u00f6rfler, Csaba Szepevari, Miroslav Krstic, Michael Muehlebach<\/strong><\/p>\n<ul><li><h4>Presentation and Speaker Bios<\/h4><p>Reinforcement learning and control theory are two adjacent scientific fields<br>that focus on optimizing the controller of unknown dynamical systems using feedback.\u00a0While both fields have common roots in dynamic programming, they have evolved with\u00a0distinct methodologies, goals, and cultures. Despite decades of mutual influence, a\u00a0significant gap persists between the two communities.<\/p> <p>This tutorial introduces adaptive\u00a0control, actor-critic reinforcement algorithms, and a new original way to combine\u00a0these two paradigms for data-driven decision making on a classical locomotion control\u00a0problem. Our aim is to provide keys to understand the core differences between both\u00a0approaches, and insights to help experts in each field better understand and engage\u00a0with the tools and approaches of the other.<\/p> <strong>Claire Vernade<\/strong> is Full Professor of Foundations of Machine Learning at the University of Technology Nuremberg (UTN). Her research focuses on sequential decision making under uncertainty, spanning reinforcement learning, online learning, and statistical machine learning. She develops theoretically grounded learning and decision-making algorithms for adaptive and interactive systems, with an emphasis on bridging mathematical foundations and scalable machine learning methods. Before joining UTN in 2025, she was a Group Leader at the University of T\u00fcbingen and a Senior Research Scientist at Google DeepMind. She is the recipient of an Emmy Noether Programme grant and an ERC Starting Grant. <br><strong>Onno Eberhard<\/strong> is a Ph.D. student in Computer Science at the Max Planck <br>Institute for Intelligent Systems and the University of T\u00fcbingen. He holds an M.Sc. in Machine Learning from the University of T\u00fcbingen and a B.Sc. in Electrical Engineering from the University of Duisburg-Essen, and has gained professional experience at Google Research and Siemens.\u00a0His research focuses on the theoretical foundations of reinforcement learning, particularly regarding <br>partially observable environments, recurrent memory, and its intersections with control theory. <br><strong>Martha White<\/strong> is a Professor of Computing Science at the University of Alberta and a Fellow of Amii, which is one of the top machine learning centres in the world. She holds a Canada CIFAR AI Chair, a Tier 2 Canada Research Chair in Reinforcement Learning, received IEEE\u2019s \u201cAIs 10 to Watch: The Future of AI\u201d award in 2020 and was inducted into the College of New Scholars by the Royal Society of Canada in 2024. She has authored more than 80 papers in top journals and conferences. Martha is an associate editor for JMLR and TMLR, server on the RLC board and has served as co-program chair for ICLR and for RLC. Her research focus is on developing reinforcement learning algorithms that learn to adapt continually, with a focus on process control and more sustainable systems. <br><strong>Florian D\u00f6rfler<\/strong> is a Professor at the Automatic Control Laboratory at ETH Z\u00fcrich. He received his Ph.D. degree in Mechanical Engineering from the University of California at Santa Barbara in 2013. From 2013 to 2014 he was an Assistant Professor at the University of California Los Angeles. His research interests are centered around automatic control, system theory, optimization, and learning. His particular foci are on network systems, data-driven settings, and applications to power systems. He is a recipient of the R\u00f6ssler Prize, the highest scientific award at ETH Z\u00fcrich across all disciplines, as well as the distinguished career awards by IFAC (Manfred Thoma Medal) and EUCA (European Control Award). He and his team received best paper distinctions in the top venues of control, power systems, power electronics, circuits and systems. <br><strong>Csaba Szepesv\u00e1ri<\/strong> is a Canada CIFAR AI Chair, Professor of Computing Science <br>at the University of Alberta, and Team Lead for the Foundations team at DeepMind. He is a Fellow of the Association for the Advancement of Artificial Intelligence and an IEEE Fellow. He received his PhD in 1999 from J\u00f3zsef Attila University in Szeged, Hungary, in probability and statistics. His research <br>advances the foundations of learning-based artificial intelligence, especially reinforcement learning, bandit algorithms, and planning under uncertainty. He is the author\u00a0or co-author of three books, including Bandit Algorithms, published by Cambridge University Press in 2020. He is also the co-inventor of UCT, an algorithm that helped ignite the modern development of Monte Carlo tree search and made simulation-based planning a central tool in game AI and decision making under uncertainty. <br><strong>Miroslav Krstic<\/strong> is a professor and serves as senior associate vice chancellor for research at UC San Diego. He is the recipient of the IEEE Brockett Award and Bode Prize, ASME Oldenburger Medal, SIAM Reid Prize, Bellman Award, and other recognitions, including the Chestnut prize and several IFAC TC awards. He is a member of the Serbian Academy of Sciences and Arts, Academia Europaea, fellow of IEEE, IFAC, SIAM, ASME, AIAA, and other societies, and Fellow-Ambassador of CNRS. Krstic is the current editor-in-chief of IEEE Transactions on Automatic Control, a former EiC of Systems &amp; Control Letters, and former senior editor in Automatica. He is a coauthor of 19 books and several hundred papers on various nonlinear, adaptive, and infinite-dimensional control subjects. <br><strong>Michael M\u00fchlebach<\/strong> leads the research group learning and dynamical systems at the Max Planck Institute for Intelligent Systems in T\u00fcbingen, Germany. His group conducts fundamental research in machine learning, reinforcement learning, and large-scale optimization. He won numerous awards including an Emmy Noether and Branco Weiss fellowship, as well as an ETH Medal and the HILTI prize for innovative research. He is also a member of the editorial board of Foundations and Trends in Machine Learning.<\/li><\/ul>\n<p>Reinforcement learning and control theory are two adjacent scientific fields<br>that focus on optimizing the controller of unknown dynamical systems using feedback.\u00a0While both fields have common roots in dynamic programming, they have evolved with\u00a0distinct methodologies, goals, and cultures. Despite decades of mutual influence, a\u00a0significant gap persists between the two communities.<\/p> <p>This tutorial introduces adaptive\u00a0control, actor-critic reinforcement algorithms, and a new original way to combine\u00a0these two paradigms for data-driven decision making on a classical locomotion control\u00a0problem. Our aim is to provide keys to understand the core differences between both\u00a0approaches, and insights to help experts in each field better understand and engage\u00a0with the tools and approaches of the other.<\/p> <strong>Claire Vernade<\/strong> is Full Professor of Foundations of Machine Learning at the University of Technology Nuremberg (UTN). Her research focuses on sequential decision making under uncertainty, spanning reinforcement learning, online learning, and statistical machine learning. She develops theoretically grounded learning and decision-making algorithms for adaptive and interactive systems, with an emphasis on bridging mathematical foundations and scalable machine learning methods. Before joining UTN in 2025, she was a Group Leader at the University of T\u00fcbingen and a Senior Research Scientist at Google DeepMind. She is the recipient of an Emmy Noether Programme grant and an ERC Starting Grant. <br><strong>Onno Eberhard<\/strong> is a Ph.D. student in Computer Science at the Max Planck <br>Institute for Intelligent Systems and the University of T\u00fcbingen. He holds an M.Sc. in Machine Learning from the University of T\u00fcbingen and a B.Sc. in Electrical Engineering from the University of Duisburg-Essen, and has gained professional experience at Google Research and Siemens.\u00a0His research focuses on the theoretical foundations of reinforcement learning, particularly regarding <br>partially observable environments, recurrent memory, and its intersections with control theory. <br><strong>Martha White<\/strong> is a Professor of Computing Science at the University of Alberta and a Fellow of Amii, which is one of the top machine learning centres in the world. She holds a Canada CIFAR AI Chair, a Tier 2 Canada Research Chair in Reinforcement Learning, received IEEE\u2019s \u201cAIs 10 to Watch: The Future of AI\u201d award in 2020 and was inducted into the College of New Scholars by the Royal Society of Canada in 2024. She has authored more than 80 papers in top journals and conferences. Martha is an associate editor for JMLR and TMLR, server on the RLC board and has served as co-program chair for ICLR and for RLC. Her research focus is on developing reinforcement learning algorithms that learn to adapt continually, with a focus on process control and more sustainable systems. <br><strong>Florian D\u00f6rfler<\/strong> is a Professor at the Automatic Control Laboratory at ETH Z\u00fcrich. He received his Ph.D. degree in Mechanical Engineering from the University of California at Santa Barbara in 2013. From 2013 to 2014 he was an Assistant Professor at the University of California Los Angeles. His research interests are centered around automatic control, system theory, optimization, and learning. His particular foci are on network systems, data-driven settings, and applications to power systems. He is a recipient of the R\u00f6ssler Prize, the highest scientific award at ETH Z\u00fcrich across all disciplines, as well as the distinguished career awards by IFAC (Manfred Thoma Medal) and EUCA (European Control Award). He and his team received best paper distinctions in the top venues of control, power systems, power electronics, circuits and systems. <br><strong>Csaba Szepesv\u00e1ri<\/strong> is a Canada CIFAR AI Chair, Professor of Computing Science <br>at the University of Alberta, and Team Lead for the Foundations team at DeepMind. He is a Fellow of the Association for the Advancement of Artificial Intelligence and an IEEE Fellow. He received his PhD in 1999 from J\u00f3zsef Attila University in Szeged, Hungary, in probability and statistics. His research <br>advances the foundations of learning-based artificial intelligence, especially reinforcement learning, bandit algorithms, and planning under uncertainty. He is the author\u00a0or co-author of three books, including Bandit Algorithms, published by Cambridge University Press in 2020. He is also the co-inventor of UCT, an algorithm that helped ignite the modern development of Monte Carlo tree search and made simulation-based planning a central tool in game AI and decision making under uncertainty. <br><strong>Miroslav Krstic<\/strong> is a professor and serves as senior associate vice chancellor for research at UC San Diego. He is the recipient of the IEEE Brockett Award and Bode Prize, ASME Oldenburger Medal, SIAM Reid Prize, Bellman Award, and other recognitions, including the Chestnut prize and several IFAC TC awards. He is a member of the Serbian Academy of Sciences and Arts, Academia Europaea, fellow of IEEE, IFAC, SIAM, ASME, AIAA, and other societies, and Fellow-Ambassador of CNRS. Krstic is the current editor-in-chief of IEEE Transactions on Automatic Control, a former EiC of Systems &amp; Control Letters, and former senior editor in Automatica. He is a coauthor of 19 books and several hundred papers on various nonlinear, adaptive, and infinite-dimensional control subjects. <br><strong>Michael M\u00fchlebach<\/strong> leads the research group learning and dynamical systems at the Max Planck Institute for Intelligent Systems in T\u00fcbingen, Germany. His group conducts fundamental research in machine learning, reinforcement learning, and large-scale optimization. He won numerous awards including an Emmy Noether and Branco Weiss fellowship, as well as an ETH Medal and the HILTI prize for innovative research. He is also a member of the editorial board of Foundations and Trends in Machine Learning.","_links":{"self":[{"href":"https:\/\/meetings.informs.org\/wordpress\/annual\/wp-json\/wp\/v2\/pages\/12076","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/meetings.informs.org\/wordpress\/annual\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/meetings.informs.org\/wordpress\/annual\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/meetings.informs.org\/wordpress\/annual\/wp-json\/wp\/v2\/users\/46"}],"replies":[{"embeddable":true,"href":"https:\/\/meetings.informs.org\/wordpress\/annual\/wp-json\/wp\/v2\/comments?post=12076"}],"version-history":[{"count":230,"href":"https:\/\/meetings.informs.org\/wordpress\/annual\/wp-json\/wp\/v2\/pages\/12076\/revisions"}],"predecessor-version":[{"id":12959,"href":"https:\/\/meetings.informs.org\/wordpress\/annual\/wp-json\/wp\/v2\/pages\/12076\/revisions\/12959"}],"wp:attachment":[{"href":"https:\/\/meetings.informs.org\/wordpress\/annual\/wp-json\/wp\/v2\/media?parent=12076"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}