Mathematics Colloquia and Seminars

Return to Colloquia & Seminar listing

Description

With the ever-growing volume of data, understanding the computational aspects of statistical inference has become increasingly critical. In this talk, we focus on the computational aspects of hypothesis selection, a fundamental problem in learning theory and statistics. The task is to select a distribution from a finite set of candidate distributions that best matches the underlying distribution generating the dataset. We will examine the hypothesis selection problem under computational constraints, exploring how to achieve near-linear-time algorithms with optimal accuracy. Additionally, we discuss methods for designing differentially private algorithms that attain optimal accuracy while running in near-optimal time complexity.