Debiased Contrastive Learning

By Ching-Yao Chuang et al
Published on Oct. 21, 2020
Read the original document by opening this link in a new tab.

Table of Contents

1 Introduction
2 Related Work
3 Setup and Sampling Bias in Contrastive Learning
4 Debiased Contrastive Loss
5 Experiments
5.1 CIFAR10 and STL10
5.2 ImageNet-100
5.3 Sentence Embeddings
5.4 Reinforcement Learning

Summary

Debiased Contrastive Learning is a paper that introduces a new objective for self-supervised representation learning. It corrects for the sampling bias of negative examples, improving performance in various domains such as vision, language, and reinforcement learning. The proposed debiased contrastive loss outperforms the state-of-the-art methods. The paper provides theoretical analysis and experimental results demonstrating the effectiveness of the new objective.
×
This is where the content will go.