AI Paper Atlas

Research map / Graph learning / Graph self-supervised learning

Label-invariant Augmentation for Semi-Supervised Graph Classification

Han Yue, Chunhui Zhang, Chuxu Zhang, Hongfu Liu

NeurIPS 2022 · Graph learning › Graph self-supervised learning · approach: Augmentation-free graph learning · 9 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
remove negative pairs from the contrastive loss (positive-pair-only) (ablation)GLA · four graph datasets, semi-supervisedaccuracyImproved
choose the hardest label-invariant augmented candidate instead of random or easiest (ablation)GLA · four graph datasets, 50% labelsaccuracyImproved
filter augmentations that violate label invariance (GraphCL+Label-Invariant) (ablation)GraphCL · four graph datasets, 50% labelsaccuracyImproved
augment in representation space rather than on raw graphs (ablation)GLA · four graph datasets, 50% labelsaccuracyImproved
vary perturbation magnitude eta (ablation)GLA · four graph datasets, 50% labelsaccuracyNo clear effect
GLA vs JOAOv2 at 30% label ratioJOAOv2 · average over 8 datasets, 30% labelsaccuracy: 76.41 -> 76.25Hurt
GLA vs best baseline at 50% labelsJOAOv2 · average over 8 datasets, 50% labelsaccuracy: 77.53 -> 77.85Improved
GLA vs best baseline at 70% labelsJOAOv2 · average over 8 datasets, 70% labelsaccuracy: 78.26 -> 78.73Improved
GLA vs JOAOv2 on GITHUBJOAOv2 · GITHUB, 50% labelsaccuracy: 68.97 -> 66.59Hurt

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