Research map / Graph learning / Graph self-supervised learning
S3GC: Scalable Self-Supervised Graph Clustering
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 S3GC instead of the best baseline on large graphs | graph clustering · ogbn-arxiv | NMI: 0.412 -> 0.463 | Helped |
| use S3GC instead of DGI on 111M-node graph | graph clustering · ogbn-papers100M | NMI: 0.416 -> 0.453 | Helped |
| use S3GC instead of MVGRL on small graphs | graph clustering · Cora | NMI: 0.608 -> 0.588 | Hurt |
| use both graph/attribute encoder and learnable embedding instead of GCN only (ablation) | S3GC · Reddit | NMI: 0.777 -> 0.807 | Helped |
| add node attribute GCN encoder to learnable embedding I (ablation) | S3GC · Reddit | NMI: 0.808 -> 0.807 | No clear effect |
| add learnable embedding I to GCN encoder (ablation) | S3GC · ogbn-arxiv | NMI: 0.460 -> 0.463 | Helped |
| add node attribute GCN encoder to learnable embedding I (ablation) | S3GC · ogbn-arxiv | NMI: 0.444 -> 0.463 | Helped |
| use random-walk positives instead of 1-hop neighbour positives (ablation) | S3GC with I encoder · ogbn-arxiv | NMI: 0.252 -> 0.444 | Helped |
| use SimCLR-style loss instead of Node2vec loss with same encoder and sampler (ablation) | S3GC I · ogbn-arxiv | NMI: 0.406 -> 0.444 | Helped |
| add node attributes to learnable embedding encoder (ablation) | S3GC-I · SBM weak graph weak features | NMI: 0.64 -> 0.71 | Helped |
| use shorter random walks (~5) rather than longer (ablation) | S3GC · several datasets | NMI | Helped |
Closest papers by meaning
- Every Node Is Different: Dynamically Fusing Self-Supervised Tasks for Attributed Graph Clustering (AAAI 2024 · similarity 0.89)
- DGCLUSTER: A Neural Framework for Attributed Graph Clustering via Modularity Maximization (AAAI 2024 · similarity 0.89)
- THESAURUS: Contrastive Graph Clustering by Swapping Fused Gromov-Wasserstein Couplings (AAAI 2025 · similarity 0.89)
- Revisiting Self-Supervised Heterogeneous Graph Learning from Spectral Clustering Perspective (NeurIPS 2024 · similarity 0.88)
- S3GCL: Spectral, Swift, Spatial Graph Contrastive Learning (ICML 2024 · similarity 0.87)
- Deep Contrastive Graph Learning with Clustering-Oriented Guidance (AAAI 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.