Robust Machine Learning via Privacy/Rate-Distortion Theory
This addresses the problem of adversarial robustness in machine learning for practitioners, but it appears incremental as it builds on existing information-theoretic frameworks without introducing a new paradigm.
The paper tackles the vulnerability of deep neural networks to adversarial examples by connecting optimal robust learning to the privacy-utility tradeoff, a generalization of rate-distortion theory, showing that the saddle point in this game can be found via maximum conditional entropy.
Robust machine learning formulations have emerged to address the prevalent vulnerability of deep neural networks to adversarial examples. Our work draws the connection between optimal robust learning and the privacy-utility tradeoff problem, which is a generalization of the rate-distortion problem. The saddle point of the game between a robust classifier and an adversarial perturbation can be found via the solution of a maximum conditional entropy problem. This information-theoretic perspective sheds light on the fundamental tradeoff between robustness and clean data performance, which ultimately arises from the geometric structure of the underlying data distribution and perturbation constraints.