AI Center Seminar - AI Fundamentals series - Prof. Zafer Doğan

The talk is jointly organized by the EPFL AI Center as part of the AI fundamentals seminar series.

Hosting professor: Prof. Pascal Frossard, EPFL AI Center Co-Academic Director

Title:
Learning Dynamics of Generative Models in High-Dimensional (Subspace) Learning

Abstract:
Understanding the training behavior of generative models in high-dimensional settings is crucial for improving their generalization and stability. In this talk, I will present a theoretical analysis of the learning dynamics of overparameterized generative models trained via stochastic gradient descent, focusing on the emergence of low-dimensional structure and spectral bias. I’ll also discuss implications for generalization and connections with recent empirical phenomena observed in deep learning.

Bio:
Zafer Doğan is an assistant professor at Koç University in Istanbul. He received his PhD from EPFL in 2015 under the supervision of Prof. Dimitri Van De Ville and Prof. Thierry Blu. Before joining Koç University, he was a postdoctoral researcher at Harvard University with Prof. Yue Lu. His research focuses on the theoretical foundations of machine learning, with particular emphasis on high-dimensional learning, generalization, and overparameterized models.