Seminar: Chance-Constrained Programming with Exogenous and Decision-Dependent Uncertainty
Thursday, November 14, 2019
4:00 p.m.-5:00 p.m.
DeWalt Seminar Room (EGR 2164), Glenn L. Martin Hall
Miguel Lejeune, Professor of Decision Sciences, George Washington University, Professor of Electrical and Computer Engineering, SEAS
Title: Chance-Constrained Programming with Exogenous and Decision-Dependent Uncertainty
Seminar Abstract: We study a class of joint chance-constrained stochastic problems with decision-dependent and exogenous uncertainty. A coupling function models the relationship between decision and decision-dependent random variables. We review the types of decision-dependent uncertainty problems and the use of coupling functions in a variety of contexts and discuss the importance of properly modeling decision-dependent uncertainty. We propose reformulations equivalent to the general chance-constrained problem with decision-dependent service uncertainty context. We design a data-driven algorithmic framework which includes the derivation of convex integer relaxation problems and tight bounds, the use of new multiterm convexification methods, and the design of a nonlinear branch-and-bound algorithm featuring a new branching rule. Computational experiments based on real-life data and assessing the scalability and computational efficiency of the method will be presented.