Robust Estimation and Generative Adversarial Nets

TL;DR

This paper links $f$-GANs with depth functions to develop computationally feasible, statistically optimal robust estimators for location under contamination models, demonstrated through theory and experiments.

Contribution

It establishes a novel connection between $f$-GANs and depth functions, enabling the use of GAN training tools for robust estimation.

Findings

Discriminator network structures yield optimal robust estimators.

The approach works for Gaussian and elliptical distributions.

Experimental results confirm theoretical predictions.

Abstract

Robust estimation under Huber's $\epsilon$-contamination model has become an important topic in statistics and theoretical computer science. Statistically optimal procedures such as Tukey's median and other estimators based on depth functions are impractical because of their computational intractability. In this paper, we establish an intriguing connection between $f$-GANs and various depth functions through the lens of $f$-Learning. Similar to the derivation of $f$-GANs, we show that these depth functions that lead to statistically optimal robust estimators can all be viewed as variational lower bounds of the total variation distance in the framework of $f$-Learning. This connection opens the door of computing robust estimators using tools developed for training GANs. In particular, we show in both theory and experiments that some appropriate structures of discriminator networks with hidden layers in GANs lead to statistically optimal robust location estimators for both Gaussian distribution and general elliptical distributions where first moment may not exist.