A Simple Framework for Contrastive Learning of Visual Representations
T. Chen, S. Kornblith, M. Norouzi, and G. Hinton. (2020)cite arxiv:2002.05709Comment: ICML'2020. Code and pretrained models at https://github.com/google-research/simclr.
Abstract
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% 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, we achieve 85.8% top-5 accuracy,
outperforming AlexNet with 100X fewer labels.
Description
A Simple Framework for Contrastive Learning of Visual Representations
%0 Generic
%1 chen2020simple
%A Chen, Ting
%A Kornblith, Simon
%A Norouzi, Mohammad
%A Hinton, Geoffrey
%D 2020
%K cs.CV stat.ML
%T A Simple Framework for Contrastive Learning of Visual Representations
%U http://arxiv.org/abs/2002.05709
%X 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% 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, we achieve 85.8% top-5 accuracy,
outperforming AlexNet with 100X fewer labels.
@misc{chen2020simple,
abstract = {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% 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, we achieve 85.8% top-5 accuracy,
outperforming AlexNet with 100X fewer labels.},
added-at = {2021-11-01T06:23:54.000+0100},
author = {Chen, Ting and Kornblith, Simon and Norouzi, Mohammad and Hinton, Geoffrey},
biburl = {https://www.bibsonomy.org/bibtex/288abdb008f0a5fd46feedc6dcde62511/aerover},
description = {A Simple Framework for Contrastive Learning of Visual Representations},
interhash = {59d9de15891684a80c81e6ccf8217442},
intrahash = {88abdb008f0a5fd46feedc6dcde62511},
keywords = {cs.CV stat.ML},
note = {cite arxiv:2002.05709Comment: ICML'2020. Code and pretrained models at https://github.com/google-research/simclr},
timestamp = {2021-11-01T06:23:54.000+0100},
title = {A Simple Framework for Contrastive Learning of Visual Representations},
url = {http://arxiv.org/abs/2002.05709},
year = 2020
}