Research map / Graph learning / Knowledge and temporal graphs
Inductive Matrix Completion Based on Graph Neural 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 |
|---|---|---|---|
| replace target-user-and-item pooling with SumPooling (ablation) | IGMC · Flixster, Douban, YahooMusic, and ML-100K | RMSE: Flixster: 0.872±0.001 -> 0.879±0.002; Douban: 0.721±0.001 -> 0.729±0.003; YahooMusic: 19.1±0.138 -> 22.9±1.102; ML-100K: | Hurt |
| remove adjacent rating regularization (ablation) | IGMC · ML-100K | RMSE: 0.905±0.001 -> 0.908±0.001 | Hurt |
| remove adjacent rating regularization (ablation) | IGMC · Flixster and Douban | RMSE: Flixster: 0.872±0.001 -> 0.872±0.001; Douban: 0.721±0.001 -> 0.721±0.001 | No clear effect |
| remove adjacent rating regularization (ablation) | IGMC · YahooMusic | RMSE: 19.1±0.138 -> 19.2±0.140 | No clear effect |
| add user and item content features to the graph representation (ablation) | IGMC · Flixster and Douban | RMSE: Flixster: 0.872±0.001 -> 0.886±0.002; Douban: 0.721±0.001 -> 0.726±0.001 | Hurt |
| add user and item content features to the graph representation (ablation) | IGMC · YahooMusic | RMSE: 19.1±0.138 -> 19.1±0.069 | No clear effect |
| add user and item content features to the graph representation (ablation) | IGMC · ML-100K | RMSE: 0.905±0.001 -> 0.906±0.001 | Hurt |
| increase the ARR regularization weight to 0.1 (ablation) | IGMC · YahooMusic | RMSE: 19.1±0.138 -> 18.98±0.140 | Helped |
| train on ML-100K and transfer without retraining | IGMC · Flixster, Douban, and YahooMusic | RMSE: Flixster: 0.906; Douban: 0.759; YahooMusic: 20.1 | Helped |
| use IGMC instead of GC-MC as training ratings become sparse | IGMC and GC-MC · ML-1M at sparsity ratios below 20% | RMSE | Helped |
Closest papers by meaning
- Geometric Matrix Completion with Recurrent Multi-Graph Neural Networks (NeurIPS 2017 · similarity 0.90)
- Provable Non-linear Inductive Matrix Completion (NeurIPS 2019 · similarity 0.88)
- Partial Matrix Completion (NeurIPS 2023 · similarity 0.87)
- Multi-Label Ranking Loss Minimization for Matrix Completion (AAAI 2025 · similarity 0.84)
- A Geometric Approach to Personalized Recommendation with Set-Theoretic Constraints Using Box Embeddings (ICML 2025 · similarity 0.83)
- Feature-Structure Adaptive Completion Graph Neural Network for Cold-start Recommendation (AAAI 2025 · similarity 0.82)
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.