Speaker: Tianyu Wang Duke University
Time: 2024-04-26 14:00-2024-04-26 15:00
Venue: C19-1

Abstract:

Learning in non-Euclidean spaces is increasingly important due to the prevalence of complex data structures and the growing dimensionality of modern problems. In this talk, we address these challenges from two perspectives. First, we explore replacing Euclidean distance with a general metric, introducing a new narrowing algorithm for bandit problems in metric spaces. Second, we present Hessian estimation methods that are robust to coordinate specification. This approach enables efficient recovery of low-rank matrices without relying on incoherence assumptions. 

Short Bio:

Tianyu Wang obtained his PhD from Duke University，and his Bachelor's degree from the Hong Kong University of Science and Technology. Dr. Wang's research focuses on zeroth-order optimization and bandit problems in metric spaces.