Location: Gilman 132
When: February 2nd at 1:30 p.m.
Title: Model uncertainty and robust control, and costly information acquisition
Abstract: Model uncertainty and information acquisition are of central interest both from the theoretical and the applied point of view. In my talk, I will provide an overview of my work on these two topics. For the former, I will present my work on an end-to-end learning framework that prescribes a non-parametric policy with certified robustness, provable optimality, and efficient implementation in feature-based decision-making problems, using Wasserstein distributionally robust optimization. For the latter, I will discuss how to make optimal decisions under costly information acquisition, a subject that has not been adequately studied in stochastic settings. The underlying model gives rise to optimal stopping problems with free boundaries that depend on the irreversibility or not of the decisions of the optimizer.