[CS] Xiao Zhang Dissertation Defense/Apr 16, 2025

[via cs]

This is an announcement of Xiao Zhang's Dissertation Defense.
===============================================
Candidate: Xiao Zhang

Date: Wednesday, April 16, 2025

Time: 11 am CST

Remote Location: https://urldefense.com/v3/__https://uchicago.zoom.us/j/6432681009?pwd=bUllY1JOVE9objEwUE5QMkIySjUrZz09__;!!BpyFHLRN4TMTrA!50nQosWfq0JmeAXCuOvS3jaUYH46zR6UjbXgz-Tu6QeCbIaoJ8dreM0ikzm4fLzKnVvU5pc2euF2Pzxb-lBfug$

Location: JCL 280 

Title: On the Symbiosis of Generative Models and Representation Learning

Abstract: Generative modeling and representation learning are fundamental to modern machine learning and computer vision. A high-quality generative model that produces realistic images and videos must capture complex visual structures and patterns. This necessity establishes an intrinsic connection between generative modeling and representation learning: generative models develop rich internal representations to capture high-dimensional data distributions effectively. Conversely, understanding the internal mechanisms of generative models through representation learning provides insights into improving them. This talk will focus on the that bidirectional connection. Specifically, I will explore two core questions: (1) How can we improve the representation learning capabilities of generative models? (2) Can we leverage representations to enhance generative modeling? By addressing these questions, I aim to highlight the necessity of designing, analyzing, and advancing generative models through the lens of representation learning.

Advisors: Michael Maire

Committee Members: Michael Maire, Rebecca Willett, David Forsyth, Anand Bhattad, and Greg Shakhnarovich.