Speaker:

Yiding Jiang

Talk Title:

The Current and New Frontiers of Generalization

Date and Location:

Thursday, February 27, 2025

Bahen Centre for Information Technology, BA 3200

This lecture is open to the public. No registration is required, but space is limited.

Abstract:

Generalization is a fundamental property of machine learning problems. In the first part of this talk, I will examine the limitations of current theories for understanding generalization in deep neural networks and present new approaches for evaluating generalization error. In the second part, I will discuss how effective exploration strategies can improve generalization across diverse scenarios—including reinforcement learning in changing environments and language model pretraining.

About Yiding Jiang:

Yiding Jiang is a fifth-year PhD student at the Machine Learning Department of Carnegie Mellon University, advised by Zico Kolter. His research is focused on foundations of deep learning and reinforcement learning with a focus on generalization and exploration. His contributions have been recognized with the Google PhD Fellowship.