A Simple Framework for Contrastive Learning of Visual Representations

02/13/2020 βˆ™ by Ting Chen, et al. βˆ™ 13 βˆ™

This paper presents SimCLR: a simple framework for contrastive learning of visual representations. We simplify recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank. In order to understand what enables the contrastive prediction tasks to learn useful representations, we systematically study the major components of our framework. We show that (1) composition of data augmentations plays a critical role in defining effective predictive tasks, (2) introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and (3) contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning. By combining these findings, we are able to considerably outperform previous methods for self-supervised and semi-supervised learning on ImageNet. A linear classifier trained on self-supervised representations learned by SimCLR achieves 76.5 top-1 accuracy, which is a 7 state-of-the-art, matching the performance of a supervised ResNet-50. When fine-tuned on only 1 outperforming AlexNet with 100X fewer labels.



Vaclav Kosar βˆ™

Listen to the paper here https://youtu.be/yeR7EJ3Cq-g πŸ‘‚πŸ“°πŸ€“ To achieve superiority on self-supervised ImageNet this paper proposes a simple contrastive framework: A Simple Framework for Contrastive Learning of Visual Representations


page 4

page 7

page 13

page 16

This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.