Research map / Probabilistic and scientific modeling / Molecules and proteins
CataBEEM: Integrating Latent Interaction Categories in Node-wise Community Detection Models for Network Data
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 CataBEEM (latent interaction categories) instead of DC-SBM for node-level community detection | TalkLife temporal stability · TalkLife six months and monthly | L2 block-pair probability difference (lower better): 0.48 vs 0.70 (six months); 0.38 vs 0.59 (monthly) | Helped |
| Use CataBEEM instead of baselines for interaction prediction | TalkLife coverage ratio prediction · TalkLife 12 time splits | CPR100_10 (higher better): best in 10 of 12 splits; lower than DCSBM in Apr and Sep, and in Jun lower than E2 | Mixed |
| Incorporate latent interaction categories (CataBEEM) vs. edge-exchangeable model without categories | TalkLife coverage ratio prediction · TalkLife six months | CPR100_10 (higher better): 2.11 vs 2.06 (E2) | Helped |
| Add interaction-wise prior from LDA topics over post text (edge prior) to the model (ablation) | Synthetic Setting 3 (Hard) · Simulation category recovery | L2 norm of category labels | Helped |
| Add interaction-wise prior to the model (ablation) | Synthetic block recovery · Simulation Setting 3 (Hard) | L2 norm of block assignments | Helped |
| Better initialization (lower perturbation delta) of block and category labels (ablation) | CataBEEM variational EM · Synthetic simulations | L2 norm of block and category labels | Helped |
| Increase node degree cutoff for block label evaluation | CataBEEM block recovery · Synthetic simulations | L2 norm of block assignments | Helped |
| Use CataBEEM with DGLFRM GNN comparison | TalkLife coverage ratio prediction · TalkLife Jan | CPR100_10 (higher better): 2.84 vs 1.092 | Helped |
Closest papers by meaning
- Semantic Community Identification in Large Attribute Networks (AAAI 2016 · similarity 0.88)
- Joint Identification of Network Communities and Semantics via Integrative Modeling of Network Topologies and Node Contents (AAAI 2017 · similarity 0.88)
- Contextual Stochastic Block Models (NeurIPS 2018 · similarity 0.87)
- Learning Persistent Community Structures in Dynamic Networks via Topological Data Analysis (AAAI 2024 · similarity 0.87)
- Harnessing Multiple Correlated Networks for Exact Community Recovery (NeurIPS 2024 · similarity 0.87)
- Community-Aware Variational Autoencoder for Continuous Dynamic Networks (AAAI 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.