BEGIN:VCALENDAR VERSION:2.0 PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4// BEGIN:VEVENT UID:20250726T212058EDT-2618vAEt5D@132.216.98.100 DTSTAMP:20250727T012058Z DESCRIPTION:Virtual Informal Systems Seminar (VISS) Centre for Intelligent Machines (CIM) and Groupe d'Etudes et de Recherche en Analyse des Decision s (GERAD)\n\nZoom Link\n Meeting ID: 910 7928 6959        \n Passcode: VISS \n \n Speaker: Jayakumar Subramanian\, Senior Research Scientist\,  Media an d Data Science Research Lab at Adobe (India)\n \n Abstract:\n I will talk abo ut two papers in my talk in addition to giving a brief overview of researc h at the Media and Data Science Research Lab at Adobe India. The first par t of my talk covers reinforcement Learning (RL) in sequential estimation a nd prediction problems in healthcare. In practice\, successful RL relies o n informative latent states derived from sequential observations to develo p optimal treatment strategies. How best to construct such states in a hea lthcare setting is an open question. In this work\, we perform an empirica l study of several information encoding architectures using data from sept ic patients in the MIMIC-III dataset to form representations of a patient state. We find that sequentially formed state representations (including t he approximate information state approach)  facilitate effective policy le arning in batch settings\, validating a more thoughtful approach to repres entation learning that remains faithful to the sequential and partial natu re of healthcare data. The second part of my talk covers use of agent base d models (ABM) and machine learning frameworks for inverse generative soci al science problems. Conventional ABM frameworks are inefficient for large populations and do not differentiate between agent transition modeling an d agent behavior modeling\, which makes behavior learning in these framewo rks challenging.  To overcome these problems\, we have developed the DeepA BM framework which takes a network-centric functional architecture and is built using the concepts of graph convolutional neural networks from deep learning frameworks. Using graph convolutional networks has enabled the fo llowing key benefits in DeepABM: i) scale ABMs to large agent populations in real-time\, ii) run ABMs efficiently on GPUs\, and iii) enable more eff icient calibration of ABMs using gradient-based supervised machine learnin g instead of the status-quo randomized search methods.\n \n Bio: \n Jayakumar is a Senior Research Scientist at the Media and Data Science Research Lab at Adobe\, India. His research interests include reinforcement learning i n single and multi-agent systems. He has a Ph.D. in reinforcement learning in partially observed and multi-agent systems from 51³Ô¹ÏÍø and dual degrees (Bachelor + Master) in Aerospace Engineering from the Indian Institute of Technology\, Bombay.\n DTSTART:20210625T140000Z DTEND:20210625T150000Z LOCATION:CA\, ZOOM SUMMARY:Two Reinforcement Learning Problems from Healthcare and Generative Social Science URL:/cim/channels/event/two-reinforcement-learning-pro blems-healthcare-and-generative-social-science-331634 END:VEVENT END:VCALENDAR