Prof. Farzan Farnia

Department of Computer Science and Engineering, The Chinese University of Hong Kong, Hong Kong

Topic: Information Theory and Adversarial Machine Learning

Abstract:

Adversarial deep learning frameworks including generative adversarial networks (GANs) and adversarial training (AT) methods have achieved great success in various machine learning tasks. However, the current understanding of the factors driving their success remains largely inadequate. In these series of lectures, we present several information theoretic frameworks to analyze adversarial machine learning algorithms and discuss how these frameworks can provide a better understanding of GANs and AT methods. Specifically, we show that GANs and AT methods can be viewed as the optimization of information measures and then leverage information theory to design adversarial deep learning algorithms with improved stability and interpretability properties. Furthermore, we present optimization methods for optimizing information measures and study the existence and computation of Nash equilibria in adversarial machine learning games. Finally, we discuss efficient implementations of information theory-based adversarial learning algorithms and evaluate their performance in application to standard image recognition datasets.

Bio:

Farzan Farnia is an Assistant Professor of Computer Science and Engineering at The Chinese University of Hong Kong. Prior to joining CUHK, he was a postdoctoral research associate at the Laboratory for Information and Decision Systems, Massachusetts Institute of Technology, from 2019–2021. He received his master’s and PhD degrees in electrical engineering from Stanford University and his bachelor’s degree in electrical engineering and mathematics from Sharif University of Technology. At Stanford, he was a graduate research assistant at the Information Systems Laboratory advised by David Tse. Farzan’s research interests span statistical learning theory, information theory, and convex optimization.