※ Zomm 병행: https://snu-ac-kr.zoom.us/j/2473239867

Modern optimization problems often involve uncertain model parameters, but the probability distribution quantifying the uncertainty is known ambiguously. Motivated by this, distributionally robust optimization frameworks are developed to provide a systematic way of hedging against the distributional ambiguity. In this talk, we focus on chance-constrained optimization, where the decision-maker needs to find a solution satisfying given constraints with high probability while optimizing the objective. We present a mixed-integer programming reformulation of the problem under Wasserstein ambiguity and show how discrete optimization techniques can help scale up computational efficiency. This is based on joint works with Nam Ho-Nguyen, Fatma Kilinc-Karzan, and Simge Kucukyavuz.