AI Paper Atlas

Research map / Label-efficient and robust learning / Self-supervised and contrastive

Understanding Negative Samples in Instance Discriminative Self-supervised Representation Learning

Kento Nozawa, Issei Sato

NeurIPS 2021 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Contrastive versus non-contrastive theory · 13 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Increase the number of negatives from K+1=32 to K+1=512SimCLR with ResNet-18 · CIFAR-10 validationmean classifier accuracy: 72.75 -> 80.12Improved
Increase the number of negatives from K+1=32 to K+1=512SimCLR with ResNet-18 · CIFAR-10 validationlinear classifier accuracy: 77.13 -> 84.13Improved
Increase the number of negatives from K+1=32 to K+1=512SimCLR with ResNet-18, projection head discarded for evaluation · CIFAR-10 validationlinear classifier accuracy: 82.02 -> 87.66Improved
Increase the number of negatives from K+1=128 to K+1=1024SimCLR with ResNet-18 · CIFAR-100 validationmean classifier accuracy: 32.67 -> 37.44Improved
Increase the number of negatives from K+1=128 to K+1=1024SimCLR with ResNet-18 · CIFAR-100 validationlinear classifier accuracy: 41.95 -> 46.57Improved
Increase the number of negatives from K+1=128 to K+1=1024SimCLR with ResNet-18, projection head discarded for evaluation · CIFAR-100 validationlinear classifier accuracy: 57.92 -> 59.46Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy