Timezone: »

 
Poster
What Makes for Good Views for Contrastive Learning?
Yonglong Tian · Chen Sun · Ben Poole · Dilip Krishnan · Cordelia Schmid · Phillip Isola

Thu Dec 10 09:00 AM -- 11:00 AM (PST) @ Poster Session 5 #1328
in Poster Session 6 »

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).

Author Information

Yonglong Tian (MIT)
Chen Sun (Google Research)
Ben Poole (Google Brain)
Dilip Krishnan (Google)
Cordelia Schmid (Google)
Phillip Isola (Massachusetts Institute of Technology)

More from the Same Authors