Research map / Label-efficient and robust learning / Self-supervised and contrastive
Graph Adversarial Self-Supervised Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| replace handcrafted views with adversarial views (BYOL backbone fixed) (ablation) | GraphCL-BYOL · COLLAB | accuracy (%): 73.3 -> 78.0 | Helped |
| replace handcrafted views with adversarial views (ablation) | GraphCL-BYOL · PTC-MR | accuracy (%): 60.5 -> 64.6 | Helped |
| replace handcrafted views with adversarial views (ablation) | GraphCL-BYOL · IMDB-M | accuracy (%): 51.6 -> 51.7 | No clear effect |
| replace SimCLR backbone with BYOL backbone (ablation) | GraphCL · MUTAG | accuracy (%): 86.8 -> 90.5 | Helped |
| perturb first hidden layer instead of input features (ablation) | GASSL · BBBP | ROC-AUC (%): 65.1 -> 67.1 | Helped |
| perturb first hidden layer instead of input features (ablation) | GASSL · HIV | ROC-AUC (%): 76.7 -> 76.8 | No clear effect |
| increase batch size from 128 to 1024 (ablation) | GASSL-H · BBBP | ROC-AUC (%): 70.6 -> 65.4 | Hurt |
| increase GNN layers from 2 to 5 (ablation) | GASSL · BBBP | ROC-AUC (%): 70.6 -> 63.8 | Hurt |
| increase embedding dimension from 128 to 512 (ablation) | GASSL · HIV | ROC-AUC (%): 74.8 -> 78.7 | Helped |
| increase embedding dimension from 128 to 512 (ablation) | GASSL · BBBP | ROC-AUC (%): 70.6 -> 63.8 | Hurt |
| increase ascent steps T from 1 to 3 with gradient accumulation (ablation) | GASSL-GIN · IMDB-M | accuracy (%): 51.5 ± 2.3 -> 51.7 ± 2.5 | No clear effect |
| increase ascent steps T from 1 to 3 (ablation) | GASSL-GIN · MUTAG | accuracy (%): 89.8 ± 5.8 -> 90.9 ± 7.9 | No clear effect |
Closest papers by meaning
- Unsupervised Adversarially Robust Representation Learning on Graphs (AAAI 2022 · similarity 0.90)
- Adversarial Training for Graph Neural Networks: Pitfalls, Solutions, and New Directions (NeurIPS 2023 · similarity 0.89)
- Self-supervised Adversarial Purification for Graph Neural Networks (ICML 2025 · similarity 0.88)
- Defending Graph Convolutional Networks against Dynamic Graph Perturbations via Bayesian Self-Supervision (AAAI 2022 · similarity 0.88)
- Adversarial Contrastive Graph Augmentation with Counterfactual Regularization (AAAI 2025 · similarity 0.86)
- Uncovering the Redundancy in Graph Self-supervised Learning Models (NeurIPS 2024 · similarity 0.86)
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.