Efficient Distribution Matching of Representations via Noise-Injected Deep InfoMax

TL;DR

This paper enhances Deep InfoMax by injecting noise into the encoder outputs to enable automatic distribution matching of learned representations, improving their conformity to specified priors for various downstream tasks.

Contribution

It introduces a novel noise-injection method into DIM to facilitate automatic distribution matching of representations, addressing a gap in existing SSRL techniques.

Findings

Enables learning of uniformly and normally distributed representations

Shows a moderate trade-off between downstream task performance and distribution matching quality

Validates approach on various downstream tasks

Abstract

Deep InfoMax (DIM) is a well-established method for self-supervised representation learning (SSRL) based on maximization of the mutual information between the input and the output of a deep neural network encoder. Despite the DIM and contrastive SSRL in general being well-explored, the task of learning representations conforming to a specific distribution (i.e., distribution matching, DM) is still under-addressed. Motivated by the importance of DM to several downstream tasks (including generative modeling, disentanglement, outliers detection and other), we enhance DIM to enable automatic matching of learned representations to a selected prior distribution. To achieve this, we propose injecting an independent noise into the normalized outputs of the encoder, while keeping the same InfoMax training objective. We show that such modification allows for learning uniformly and normally distributed representations, as well as representations of other absolutely continuous distributions. Our approach is tested on various downstream tasks. The results indicate a moderate trade-off between the performance on the downstream tasks and quality of DM.