What Makes for Good Views for Contrastive Learning?

TL;DR

This paper investigates how the choice of views in contrastive learning affects performance, proposing that reducing mutual information between views while preserving task-relevant info enhances learning, leading to state-of-the-art results.

Contribution

The paper introduces a theoretical and empirical framework for selecting views in contrastive learning by reducing mutual information, improving downstream task performance.

Findings

Reducing mutual information between views improves classification accuracy.

Increasing data augmentation decreases mutual information and enhances performance.

Achieved new state-of-the-art accuracy on ImageNet unsupervised pre-training.

Abstract

Contrastive learning between multiple views of the data has recently achieved state of the art performance in the field of self-supervised representation learning. Despite its success, the influence of different view choices has been less studied. In this paper, we use theoretical and empirical analysis to better understand the importance of view selection, and argue that we should reduce the mutual information (MI) between views while keeping task-relevant information intact. To verify this hypothesis, we devise unsupervised and semi-supervised frameworks that learn effective views by aiming to reduce their MI. We also consider data augmentation as a way to reduce MI, and show that increasing data augmentation indeed leads to decreasing MI and improves downstream classification accuracy. As a by-product, we achieve a new state-of-the-art accuracy on unsupervised pre-training for ImageNet classification ($73\%$ top-1 linear readout with a ResNet-50). In addition, transferring our models to PASCAL VOC object detection and COCO instance segmentation consistently outperforms supervised pre-training. Code:http://github.com/HobbitLong/PyContrast