Research map / Probabilistic and scientific modeling / Variational and generative sampling
Relational Deep Learning: A Deep Latent Variable Model for Link Prediction
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 RDL joint modeling of attributes and links instead of gRTM | RDL · citeulike-a, K=50 | link rank: 744 -> 495 | Helped |
| use RDL instead of gRTM | RDL · arXiv, K=50 | AUC (%): 86.78% -> 90.52% | Helped |
| increase RDL depth from 1 to 2 layers (ablation) | RDL (MAP) · arXiv | link rank: 5241.97 -> 2080.72 | Helped |
| increase RDL depth from 2 to 3 layers (ablation) | RDL (MAP) · arXiv | link rank: 2080.72 -> 2730.08 | Hurt |
| increase RDL depth from 2 to 3 layers (ablation) | RDL (MAP) · citeulike-a | link rank: 495.97 -> 488.41 | Helped |
| increase RDL depth from 2 to 3 layers (ablation) | RDL (MAP) · citeulike-a | AUC: 0.964 -> 0.963 | No clear effect |
| replace MAP estimation with Bayesian treatment (GVI) (ablation) | RDL · citeulike-a | link rank: 495.97 -> 473.59 | Helped |
| replace MAP estimation with Bayesian treatment (GVI) (ablation) | RDL · arXiv | link rank: 2028.72 -> 1982.84 | Helped |
| add estimated variance as BLR feature to BSDAE mean (ablation) | BSDAE+BLR · citeulike-a | link rank: 849.02 -> 761.57 | Helped |
| train jointly with links (BayesRDL) rather than BSDAE2 features plus BLR (ablation) | BSDAE2+BLR · citeulike-a | link rank: 761.57 -> 473.59 | Helped |
| Bayesian SDAE features vs plain SDAE autoencoder features, both with BLR | SDAE+BLR · citeulike-a | link rank: 992.48 -> 761.57 | Helped |
Closest papers by meaning
- Deep Generative Models for Relational Data with Side Information (ICML 2017 · similarity 0.89)
- Scalable Deep Generative Relational Model with High-Order Node Dependence (NeurIPS 2019 · similarity 0.89)
- Discriminative Nonparametric Latent Feature Relational Models with Data Augmentation (AAAI 2016 · similarity 0.88)
- RelGNN: Composite Message Passing for Relational Deep Learning (ICML 2025 · similarity 0.87)
- Position: Relational Deep Learning - Graph Representation Learning on Relational Databases (ICML 2024 · similarity 0.87)
- Learning Latent Graph Structures and their Uncertainty (ICML 2025 · similarity 0.85)
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.