[PAST EVENT] Data-driven Stochastic Optimization: A Learning-based Perspective
Speaker: Kwai Hung Henry Lam
Abstract: We discuss a statistical framework to integrate data into optimization under uncertain constraints, via the use of suitably constructed robust optimization (RO) that provides inner approximation to a benchmark chance-constrained problem. The framework is based on learning a prediction set with geometries compatible with established RO tools and validating the set to achieve finite-sample statistical guarantees on feasibility. We present and compare the performances and sample size requirements of our method with previous approaches, and demonstrate how our framework provides a platform to integrate machine learning tools into uncertain optimization driven by convoluted or high-dimensional data. This is joint work with Jeff Hong and Zhiyuan Huang.
Henry Lam is an Assistant Professor in the Department of Industrial and Operations Engineering at the University of Michigan. He received his Ph.D. in statistics from Harvard University in 2011. His research focuses on Monte Carlo methods, risk analysis and stochastic and simulation-based optimization. His work has been recognized by an NSF CAREER Award, INFORMS JFIG Competition Second Prize, INFORMS Nicholson Competition Honorable Mention Prize, and an Adobe Faculty Research Award. He is currently an Associate Editor for Operations Research and INFORMS Journal on Computing.