Learning Representations by Maximizing Mutual Information Across Views

TL;DR

This paper introduces a self-supervised learning method that maximizes mutual information across different views of data, leading to superior image representations and emergent segmentation capabilities.

Contribution

It presents a novel mutual information maximization framework for self-supervised learning that improves image representation quality and reveals segmentation behavior.

Findings

Achieves 68.1% accuracy on ImageNet with linear evaluation.

Outperforms prior methods by over 12% on ImageNet.

Emerges segmentation behavior with mixture-based representations.

Abstract

We propose an approach to self-supervised representation learning based on maximizing mutual information between features extracted from multiple views of a shared context. For example, one could produce multiple views of a local spatio-temporal context by observing it from different locations (e.g., camera positions within a scene), and via different modalities (e.g., tactile, auditory, or visual). Or, an ImageNet image could provide a context from which one produces multiple views by repeatedly applying data augmentation. Maximizing mutual information between features extracted from these views requires capturing information about high-level factors whose influence spans multiple views -- e.g., presence of certain objects or occurrence of certain events. Following our proposed approach, we develop a model which learns image representations that significantly outperform prior methods on the tasks we consider. Most notably, using self-supervised learning, our model learns representations which achieve 68.1% accuracy on ImageNet using standard linear evaluation. This beats prior results by over 12% and concurrent results by 7%. When we extend our model to use mixture-based representations, segmentation behaviour emerges as a natural side-effect. Our code is available online: https://github.com/Philip-Bachman/amdim-public.