Description
Date and Time:
July 5, 2024: 3:00 pm – 4:00 pm (JST)
Venue: Online and Open Space at the RIKEN AIP Nihonbashi office
TITLE: Data-Optimization Integration: Statistical Challenges and Recent Developments
SPEAKER: Prof. Henry Lam (Columbia University, USA)
ABSTRACT:
Many modern decision-making pipelines entail an integration of data into optimization that have spurred surging studies among the operations research community. Instead of learning good prediction models from data, the goal of these data-to-decision pipelines is to devise policies that perform well on the downstream objectives or optimality gaps. I will discuss some underlying statistical challenges and our recent attempts to address them. These include principled comparisons of data-driven optimization formulations (touching upon issues such as distributional robustness), the “positioning” of estimation within optimization (e.g., estimate-then-optimize versus integrated approaches), and validation methodologies (for effective decision selection in lieu of classical model selection).
Bio:
Henry Lam is an Associate Professor in the Department of Industrial Engineering and Operations Research at Columbia University. His research interests include Monte Carlo methods, uncertainty quantification, data-driven optimization and rare-event analysis. His works have been recognized by several venues such as the NSF CAREER Award and NSA Young Investigator Award, funding awards from industry such as Google, Adobe and JPMorgan, and other paper awards. He serves on the editorial boards of seven flagship journals in operations research, including Management Science, Operations Research, Manufacturing and Service Operations Management, and as the Area Editor in Stochastic Models and Data Science in Operations Research Letters. He is currently the Vice-Chair and Chair-Elect of the INFORMS Applied Probability Society. Henry holds a PhD degree in statistics from Harvard University and BS degree in actuarial science from the University of Hong Kong.
