Invariant Representations without Adversarial Training

TL;DR

This paper introduces a non-adversarial, information-theoretic method for learning invariant data representations, simplifying training and improving performance in fairness and generative modeling tasks.

Contribution

It proposes a novel approach that avoids adversarial training by directly optimizing an information-theoretic objective for invariant representations.

Findings

Matches or exceeds state-of-the-art adversarial methods

Simplifies training process

Improves fairness and controllable generative modeling

Abstract

Representations of data that are invariant to changes in specified factors are useful for a wide range of problems: removing potential biases in prediction problems, controlling the effects of covariates, and disentangling meaningful factors of variation. Unfortunately, learning representations that exhibit invariance to arbitrary nuisance factors yet remain useful for other tasks is challenging. Existing approaches cast the trade-off between task performance and invariance in an adversarial way, using an iterative minimax optimization. We show that adversarial training is unnecessary and sometimes counter-productive; we instead cast invariant representation learning as a single information-theoretic objective that can be directly optimized. We demonstrate that this approach matches or exceeds performance of state-of-the-art adversarial approaches for learning fair representations and for generative modeling with controllable transformations.