Research map / Graph learning / Knowledge and temporal graphs
Factorizable 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 graph-level factorization and factor-specific feature aggregation instead of a conventional GCN baseline | Synthetic dataset classification · Synthetic dataset | Micro-F1: 0.947 ± 0.003 -> 0.995 ± 0.004 | Helped |
| Use graph-level factorization to represent latent relations | Synthetic graph disentanglement · Synthetic dataset | C-Score: 0.367±0.026 -> 0.532 ± 0.044 | Helped |
| Use FactorGCN rather than DisenGCN for edge disentanglement | ZINC · ZINC | GED_E: 14.14±6.19 -> 12.72 ± 5.34 | Helped |
| Apply FactorGCN without edge-type information rather than GatedGCN_E with edge information | Molecular regression · ZINC | MAE: 0.366 ± 0.014 vs 0.363 ± 0.009 | No clear effect |
| Add the discriminator loss to FactorGCN (ablation) | FactorGCN · Synthetic dataset | downstream task and disentanglement performance | Helped |
| Increase the number of factor graphs from four to five (ablation) | FactorGCN · Synthetic classification, five factor graphs | Micro-F1: 0.995 ± 0.004 -> 0.893 ± 0.021 | Helped |
| Add the discriminator to GAT (ablation) | GAT · Synthetic dataset | Micro-F1: 0.923 ± 0.009 -> 0.928 ± 0.009 | No clear effect |
| Add the discriminator to GAT (ablation) | GAT · Synthetic dataset | GED_E: 12.59 ± 3.00 -> 12.35 ± 3.86 | No clear effect |
| Add the discriminator to GAT (ablation) | GAT · Synthetic dataset | C-Score: 0.288 ± 0.064 -> 0.274 ± 0.065 | Hurt |
| Increase the number of factors beyond the ground-truth count (ablation) | FactorGCN · Synthetic datasets with varying factor counts | classification performance | Mixed |
Closest papers by meaning
- Disentangled Graph Convolutional Networks (ICML 2019 · similarity 0.92)
- Independence Promoted Graph Disentangled Networks (AAAI 2020 · similarity 0.91)
- Beyond Graph Convolutional Network: An Interpretable Regularizer-Centered Optimization Framework (AAAI 2023 · similarity 0.89)
- Composition-based Multi-Relational Graph Convolutional Networks (ICLR 2020 · similarity 0.89)
- Powerful Graph Convolutional Networks with Adaptive Propagation Mechanism for Homophily and Heterophily (AAAI 2022 · similarity 0.88)
- Disentangled Contrastive Learning on Graphs (NeurIPS 2021 · similarity 0.88)
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.