AI Paper Atlas

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

Learning by Sorting: Self-supervised Learning with Group Ordering Constraints

Nina Shvetsova, Felix Petersen, Anna Kukleva, Bernt Schiele, Hilde Kuehne

ICCV 2023 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Contrastive versus non-contrastive theory · 7 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
replace InfoNCE with GroCo lossSimCLR (ResNet50, 100 ep) · ImageNet k-NN k=20accuracy: 51.9 -> 60.5Improved
replace InfoNCE with GroCo lossSimCLR (ResNet50, 100 ep) · ImageNet linear probingTop-1 accuracy: 65.7 -> 69.2Improved
replace triplet loss with GroCo (ablation)ResNet50 100 ep · ImageNet k-NN k=20accuracy: 53.8 -> 60.5Improved
pre-order positives and negatives before the sorting network (ablation)GroCo · ImageNet k-NN k=20accuracy: 59.5 -> 60.5Improved
increase negatives from 10 to 20 at beta=1 (ablation)GroCo · ImageNet k-NN k=20accuracy: 60.5 -> 54.6Hurt
increase batch size from 256 to 1024 (ablation)GroCo · ImageNet k-NN k=20accuracy: 57.2 -> 60.5Improved
sample batches with many false negatives (ablation)GroCo · ImageNet k-NN k=20accuracy: 60.5 -> 56.9Hurt
increase number of positive views from 2 to 4 (ablation)GroCo · ImageNet k-NN k=20accuracy: 60.5 -> 61.8Improved
increase number of positive views from 2 to 4 (ablation)GroCo · ImageNet linear probingTop-1 accuracy: 69.2 -> 69.6No clear effect
use top-10 negatives without stop gradient (ablation)GroCo and InfoNCE · ImageNet 100 epochstraining stabilityHurt
GroCo vs SwAV with multi-crop augmentationSwAV · ImageNet linear probing 400 epTop-1 accuracy: 74.6 -> 73.7Hurt
plug GroCo loss into MoCo v3 with ViT-SmallMoCo v3 · ImageNet k-NN k=20accuracy: 66.8 -> 67.5Improved

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