# A Simple Framework for Contrastive Learning of Visual Representations

**Authors:** Ting Chen, Simon Kornblith, Mohammad Norouzi, Geoffrey Hinton  
**arXiv:** [2002.05709](https://arxiv.org/abs/2002.05709)  
**Date:** 2020-02-13  
**Summary page:** https://www.summarizepaper.com/en/paper/2002.05709v3/

## Key points

- SimCLR is a framework for contrastive learning of visual representations.
- It simplifies recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank.
- Composition of data augmentations plays a critical role in defining effective predictive tasks.
- Introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations.
- Contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning.
- SimCLR outperforms previous methods for self-supervised and semi-supervised learning on ImageNet, achieving 7% relative improvement over previous state-of-the-art, matching the performance of a supervised ResNet-50.
- A linear classifier trained on self-supervised representations learned by SimCLR achieves 76.5% top-1 accuracy.
- When fine-tuned on only 1% of the labels, they achieve 85.8% top-5 accuracy, outperforming AlexNet with 100X fewer labels.
- The authors provide code and pretrained models at https://github.com/google-research/simclr.
- SimCLR's contributions lie in its ability to simplify existing methods while achieving state-of-the-art results in self-supervised and semi-supervised learning on ImageNet.
- Its findings regarding data augmentation composition, learnable nonlinear transformations, and batch size/training steps can be applied more broadly to improve other contrastive learning frameworks as well.

## Summary

SimCLR is a novel framework for contrastive learning of visual representations that simplifies recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank. The authors systematically study the major components of their framework to understand what enables the contrastive prediction tasks to learn useful representations. They find that composition of data augmentations plays a critical role in defining effective predictive tasks, introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning. By combining these findings, they 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% relative improvement over previous state-of-the-art, matching the performance of a supervised ResNet-50. When fine-tuned on only 1% of the labels, they achieve 85.8% top-5 accuracy, outperforming AlexNet with 100X fewer labels. The authors provide code and pretrained models at https://github.com/google-research/simclr. SimCLR's contributions lie in its ability to simplify existing methods while achieving state-of-the-art results in self-supervised and semi-supervised learning on ImageNet. Its findings regarding data augmentation composition, learnable nonlinear transformations, and batch size/training steps can be applied more broadly to improve other contrastive learning frameworks as well.