Research map / Graph learning / GNN architectures
Graph Convolution Network based Recommender Systems: Learning Guarantee and Item Mixture Powered Strategy
Leyan Deng, Defu Lian, Chenwang Wu, Enhong Chen
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 |
|---|---|---|---|
| add Item Mixture (IMix) augmentation | NGCF 3 layer · Gowalla | Recall@20: 0.1303 -> 0.1496 | Improved |
| add IMix | NGCF 1 layer · Yelp2018 | Recall@20: 0.0419 -> 0.0492 | Improved |
| add IMix | NGCF 1 layer · Gowalla | Recall@20: 0.1308 -> 0.1451 | Improved |
| add IMix | LightGCN 3 layer · Gowalla | Recall@20: 0.1759 -> 0.1793 | Improved |
| add IMix | LightGCN 2 layer · Yelp2018 | Recall@20: 0.0564 -> 0.0621 | Improved |
| add IMix | LightGCN 3 layer · Yelp2018 | NDCG@20: 0.0492 -> 0.0527 | Improved |
| use IMix on weaker NGCF versus stronger LightGCN | IMix · Gowalla and Yelp2018 | relative improvement | Mixed |
| train LightGCN with IMix beyond 50 epochs | LightGCN-IMix · Gowalla and Yelp2018 | test Recall/NDCG over epochs | Improved |
| increase number of GCN layers L (ablation) | NGCF · Gowalla | generalization gap (Recall/NDCG@K) | Hurt |
| increase spectral norm of weight matrices (ablation) | NGCF · Gowalla | generalization gap | Hurt |
| use normalized graph (D^-1 A + I or D^-1/2 A D^-1/2 + I) instead of unnormalized A + I (ablation) | NGCF · Gowalla | generalization (Recall@20, NDCG@20) | Improved |
| swap non-linear activation (LeakyReLU, ReLU, Sigmoid, Tanh, none) (ablation) | NGCF · Gowalla | Recall@20, NDCG@20 | No clear effect |
Closest papers by meaning
- Beyond Graph Convolution: Multimodal Recommendation with Topology-aware MLPs (AAAI 2025 · similarity 0.89)
- Hgformer: Hyperbolic Graph Transformer for Collaborative Filtering (ICML 2025 · similarity 0.87)
- Spectral-Based Graph Neural Networks for Complementary Item Recommendation (AAAI 2024 · similarity 0.87)
- ContextGNN: Beyond Two-Tower Recommendation Systems (ICLR 2025 · similarity 0.86)
- Semantic Enhanced Heterogeneous Hypergraph Network for Collaborative Filtering (AAAI 2025 · similarity 0.86)
- Modality-Independent Graph Neural Networks with Global Transformers for Multimodal Recommendation (AAAI 2025 · 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.
