What Makes for Good Views for Contrastive Learning?

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

PDF Abstract NeurIPS 2020 PDF NeurIPS 2020 Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Self-Supervised Image Classification ImageNet InfoMin (ResNeXt-152) Top 1 Accuracy 75.2% # 73
Number of Params 120M # 28
Self-Supervised Image Classification ImageNet InfoMin (ResNet-50) Top 1 Accuracy 73.0% # 87
Top 5 Accuracy 91.1% # 18
Number of Params 24M # 48
Contrastive Learning imagenet-1k ResNet50 ImageNet Top-1 Accuracy 73 # 2

Methods


No methods listed for this paper. Add relevant methods here