AI Paper Atlas

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

Weak Augmentation Guided Relational Self-Supervised Learning

Mingkai Zheng, Shan You, Fei Wang, Chen Qian, Changshui Zhang, Xiaogang Wang, Chang Xu

TPAMI 2024 · Label-efficient and robust learning › Self-supervised and contrastive · approach: Data-efficient and robust self-supervision · 79 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
use weak augmentation instead of contrastive augmentation for the teacher (ablation)ReSSL · Tiny ImageNetlinear top-1: 38.66 -> 46.64Improved
use weak augmentation instead of contrastive augmentation for teacher (ablation)ReSSL · CIFAR-100linear top-1: 58.60 -> 65.03Improved
add random resized crop to no teacher augmentation (ablation)ReSSL · CIFAR-10linear top-1: 79.33 -> 90.62Improved
add color jitter, grayscale and blur on top of crop and flip for teacher (ablation)ReSSL · CIFAR-100linear top-1: 65.03 -> 58.60Hurt
add InfoNCE warm-up (ablation)ReSSL · Tiny ImageNetlinear top-1: 46.64 -> 48.22Improved
add InfoNCE warm-up (ablation)ReSSL · CIFAR-10linear top-1: 90.76 -> 89.98Hurt
remove predictor (symmetric system) with tau_s=0.1 and tau_t>=0.08 (ablation)ReSSL · CIFAR-10linear top-1: 90.20 best -> 10.00 collapsedHurt
add predictor head (ablation)ReSSL · CIFAR-10linear top-1: 90.20 -> 90.76Improved
use tau_s=0.1 instead of 0.2 (ablation)ReSSL · CIFAR-100linear top-1: 63.12 -> 65.03Improved
increase teacher temperature tau_t to 0.10 (flatter target) (ablation)ReSSL · Tiny ImageNetlinear top-1: 46.64 -> 32.28Hurt
increase memory buffer from 256 to 16384 (ablation)ReSSL · Tiny ImageNetlinear top-1: 42.80 -> 46.64Improved
increase memory buffer from 16384 to 32768 (ablation)ReSSL · Tiny ImageNetlinear top-1: 46.64 -> 45.95Hurt

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