Abstract: We consider generic stochasti c optimization problems in the presence of side information which enables a more insightful decision. The side information constitutes observable exogenous covariates that alter the conditional probability distribution of the random problem parameters. A decision-maker who adapts her decisio ns according to the observed side information solves a stochastic optimiz ation problem where the objective function is specified by the conditiona l expectation of the random cost. If the joint probability distribution i s unknown, then the conditional expectation can be approximated in a data -driven manner using the Nadaraya-Watson (NW) kernel regression. While th e emerging approximation scheme has found successful applications in dive rse decision problems under uncertainty, it is largely unknown whether th e scheme can provide any reasonable out-of-sample performance guarantees. In this talk, we establish guarantees for the generic problems by levera ging techniques from moderate deviations theory. Our analysis motivates t he use of a variance-based regularization scheme which, in general, leads to a non-convex optimization problem. We adopt ideas from distributional ly robust optimization to obtain tractable formulations. We present numer ical experiments for inventory management and wind energy commitment prob lems to highlight the effectiveness of our regularization scheme.

Bio: Grani A. Hanasusanto is an Assistant Professor of Operations Resear ch and Industrial Engineering at The University of Texas at Austin (UT). Before joining UT, he was a postdoctoral researcher at the College of Man agement of Technology at École Polytechnique Fédérale de Lausanne. He holds a Ph.D. degree in Operations Research from Imperial College London and an MSc degree in Financial Engineering from the National University o f Singapore. He is the recipient of the 2018 NSF CAREER Award. His resear ch focuses on the design and analysis of tractable solution schemes for d ecision-making problems under uncertainty, with applications in operation s management, energy systems, finance, machine learning and data analytic s.

UID:20220325T185000Z-265309@calendar.tamu.edu DTSTAMP:20220118T104928Z URL:https://calendar.tamu.edu/ise/event/265309-isen-on-data-driven-prescr iptive-analytics-with LAST-MODIFIED:20220118T164928Z X-LIVEWHALE-TYPE:events X-LIVEWHALE-ID:265309 X-LIVEWHALE-TIMEZONE:America/Chicago X-LIVEWHALE-SUMMARY:A seminar with Dr. Grani A. Hanasusanto\, assistant p rofessor at the University of Texas\, Austin. \; X-LIVEWHALE-TAGS:seminar|Seminars X-LIVEWHALE-CUSTOM-CUSTOM-ROOM-NUMBER:1034 END:VEVENT END:VCALENDAR