Research map / Graph learning / GNN architectures
When Does Self-Supervision Help Graph Convolutional Networks?
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 |
|---|---|---|---|
| Use multi-task learning rather than pretraining and finetuning with graph partitioning | GCN · PubMed | accuracy: 79.10 $\pm$ 0.21 -> 80.00 $\pm$ 0.74 | Helped |
| Add graph-partitioning self-supervision through multi-task learning | GCN · Cora | accuracy: 81.00 $\pm$ 0.67 -> 81.83 $\pm$ 0.65 | Mixed |
| Add graph-partitioning self-supervision through multi-task learning | GCN · Citeseer | accuracy: 70.85 $\pm$ 0.70 -> 71.34 $\pm$ 0.69 | Helped |
| Add node-clustering self-supervision through multi-task learning | GCN · PubMed | accuracy: 79.10 $\pm$ 0.21 -> 78.79 $\pm$ 0.36 | Hurt |
| Add graph-completion self-supervision through multi-task learning | GCN · Citeseer | accuracy: 70.85 $\pm$ 0.70 -> 71.66 $\pm$ 0.48 | Helped |
| Use pretraining and finetuning with self-supervision | GCN · PubMed | accuracy: 79.10 $\pm$ 0.21 -> 79.19 $\pm$ 0.21 | No clear effect |
| Use multi-task graph partitioning instead of self-training | GCN · Citeseer | accuracy: 71.94 $\pm$ 0.83 -> 71.34 $\pm$ 0.69 | Mixed |
| Add graph-partitioning self-supervision through multi-task learning | GAT, GIN, GMNN, and GraphMix · Cora, Citeseer, and PubMed | accuracy | Helped |
| Add node-clustering self-supervision through multi-task learning | GCN, GAT, GIN, GMNN, and GraphMix · PubMed | accuracy | Hurt |
| Add graph-completion self-supervision through multi-task learning | GraphMix · Cora | accuracy: 83.91 $\pm$ 0.63 -> 83.76 $\pm$ 0.64 | Hurt |
| Add node-clustering self-supervision to adversarial training | GCN · Cora feature attacks | accuracy: 75.25 $\pm$ 1.26 -> 76.24 $\pm$ 0.99 | Helped |
| Add graph-partitioning self-supervision to adversarial training | GCN · Cora link attacks | accuracy: 54.58 $\pm$ 2.57 -> 56.36 $\pm$ 2.57 | Helped |
Closest papers by meaning
- Defending Graph Convolutional Networks against Dynamic Graph Perturbations via Bayesian Self-Supervision (AAAI 2022 · similarity 0.91)
- Deeper Insights Into Graph Convolutional Networks for Semi-Supervised Learning (AAAI 2018 · similarity 0.89)
- L2-GCN: Layer-Wise and Learned Efficient Training of Graph Convolutional Networks (CVPR 2020 · similarity 0.89)
- SLAPS: Self-Supervision Improves Structure Learning for Graph Neural Networks (NeurIPS 2021 · similarity 0.88)
- Uncovering the Redundancy in Graph Self-supervised Learning Models (NeurIPS 2024 · similarity 0.88)
- Exploring Correlations of Self-Supervised Tasks for Graphs (ICML 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.