Research map / Graph learning / Graph self-supervised learning
Contrastive Graph Structure Learning via Information Bottleneck for Recommendation
Chunyu Wei, Jian Liang, Di Liu, Fei Wang
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 changed | Where | Measured | Effect |
|---|---|---|---|
| replace random augmentation (SGL) with learnable IB-optimized augmentation (CGI) | LightGCN contrastive learning · Yelp2018 | NDCG@10: 0.0367 -> 0.0392 | Improved |
| replace SGL with CGI | LightGCN contrastive learning · MovieLens-1M | RECALL@10: 0.1965 -> 0.2180 | Improved |
| replace SGL with CGI | LightGCN contrastive learning · Douban | NDCG@10: 0.0912 -> 0.0991 | Improved |
| add contrastive learning with random dropout (SGL) to LightGCN | LightGCN · Yelp2018 | NDCG@10: 0.0344 -> 0.0367 | Improved |
| use only learnable node-dropping view instead of both views (ablation) | CGI · Yelp2018 | NDCG@10: 0.0392 -> 0.0369 | Hurt |
| use only learnable edge-dropping view instead of both views (ablation) | CGI · Yelp2018 | NDCG@10: 0.0392 -> 0.0379 | Hurt |
| use only learnable node-dropping view instead of both views (ablation) | CGI · MovieLens-1M | NDCG@10: 0.1979 -> 0.1934 | Hurt |
| use only learnable edge-dropping view instead of both views (ablation) | CGI · MovieLens-1M | NDCG@10: 0.1979 -> 0.1916 | Hurt |
| learnable node-dropping instead of random node dropout (ablation) | SGL-ND · Yelp2018 | NDCG@10: 0.0356 -> 0.0369 | Improved |
| learnable edge-dropping instead of random edge dropout (ablation) | SGL-ED · MovieLens-1M | NDCG@10: 0.1800 -> 0.1916 | Improved |
| replace IB contrastive objective with mutual-information maximization (GCL) or remove contrastive learning (GL) (ablation) | CGI · Yelp2018 | NDCG@10/Recall@10 | Hurt |
| add CGI augmentation to NGCF backbone | NGCF · Yelp2018 | NDCG@10: 0.0242 -> 0.0272 | Improved |
Closest papers by meaning
- Robust Graph Based Social Recommendation Through Contrastive Multi-View Learning (AAAI 2025 · similarity 0.88)
- DivGCL: A Graph Contrastive Learning Model for Diverse Recommendation (AAAI 2025 · similarity 0.88)
- Review-Enhanced Hierarchical Contrastive Learning for Recommendation (AAAI 2024 · similarity 0.87)
- Graph Disentangled Contrastive Learning with Personalized Transfer for Cross-Domain Recommendation (AAAI 2024 · similarity 0.87)
- DisCo: Graph-Based Disentangled Contrastive Learning for Cold-Start Cross-Domain Recommendation (AAAI 2025 · similarity 0.87)
- Temporal Graph Contrastive Learning for Sequential Recommendation (AAAI 2024 · similarity 0.87)
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.
