CS Colloquium: Ioannis Mitliagkas (University of Montréal) - Negative Momentum for Improved Game Dynamics
Tue, Oct 23, 2018 @ 03:30 PM - 04:50 PM
Conferences, Lectures, & Seminars
Speaker: Ioannis Mitliagkas, University of Montréal
Talk Title: Negative Momentum for Improved Game Dynamics
Series: Computer Science Colloquium
Abstract: Games generalize the single-objective optimization paradigm by introducing different objective functions for different players. Differentiable games often proceed by simultaneous or alternating gradient updates. In machine learning, games are gaining new importance through formulations like generative adversarial networks (GANs) and actor-critic systems. However, compared to single-objective optimization, game dynamics are more complex and less understood. In this talk, I will present recent research on the momentum dynamics of differentiable games. We will see an analysis of a simple differentiable game, which suggests that a negative momentum term can sometimes improve convergence. Then we will see empirical results that alternating gradient updates with a negative momentum term achieves convergence on the notoriously difficult to train saturating GANs.
This lecture satisfies requirements for CSCI 591: Research Colloquium.
Biography: Ioannis Mitliagkas is an assistant professor in the department of Computer Science and Operations Research (DIRO) at the University of Montréal, and member of MILA. Before that, he was a Postdoctoral Scholar with the departments of Statistics and Computer Science at Stanford University. He obtained his Ph.D. from the department of Electrical and Computer Engineering at The University of Texas at Austin. His research includes topics in statistical learning and inference, focusing on optimization, efficient large-scale and distributed algorithms, statistical learning theory and MCMC methods. His recent work includes methods for efficient and adaptive optimization, studying the interaction between optimization and the dynamics of large-scale learning systems as well as understanding and improving the performance of Gibbs samplers. In the past he has worked on high-dimensional streaming problems and fast algorithms and computation for large graph problems.
Host: Fei Sha
Audiences: Everyone Is Invited
Contact: Computer Science Department