Research map / Graph learning / Graph algorithms and clustering
Weighted Flow Diffusion for Local Graph Clustering with Node Attributes: an Algorithm and Statistical Guarantees
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 |
|---|---|---|---|
| Reweight graph edges using node-attribute similarity | Flow diffusion · Co-authorship networks, 20 clusters (average) | F1: 50.3 -> 54.6; +4.3 | Helped |
| Use node attributes rather than unweighted edges (ablation) | Flow diffusion · Computer Vision | F1: 37.6 -> 35.5; -2.1 | Hurt |
| Use node attributes rather than unweighted edges (ablation) | Flow diffusion · Information Theory | F1: 47.4 -> 46.9; -0.5 | Hurt |
| Use node attributes rather than unweighted edges (ablation) | Flow diffusion · Robotics | F1: 59.9 -> 59.9; 0.0 | No clear effect |
| Use node attributes rather than unweighted edges (ablation) | Flow diffusion · Co-authorship networks, most clusters | F1: 50.3 -> 54.6; +4.3 average | Mixed |
| Use node attributes when their signal is weak (below the stated useful-signal threshold) (ablation) | Flow diffusion · Synthetic SBM | F1 | Hurt |
| Increase node-attribute signal strength (ablation) | Weighted flow diffusion · Synthetic SBM | F1 | Helped |
| Set sink capacities to node degrees rather than 1 (ablation) | Weighted flow diffusion · Co-authorship networks | F1 | Helped |
| Increase internal edge probability from 0.01 to 0.03 while keeping the attributes fixed (ablation) | Weighted flow diffusion · Synthetic SBM with moderately strong attributes | Recall | Helped |
| Reduce attribute signal from 3 to 5/2 times the stated scale (ablation) | Weighted flow diffusion · Synthetic SBM, p=0.01 and q=0.002 | Recall | Hurt |
Closest papers by meaning
- Local Graph Clustering with Noisy Labels (ICLR 2024 · similarity 0.93)
- p-Norm Flow Diffusion for Local Graph Clustering (ICML 2020 · similarity 0.91)
- Capacity Releasing Diffusion for Speed and Locality (ICML 2017 · similarity 0.91)
- Optimal Graph Clustering without Edge Density Signals (NeurIPS 2025 · similarity 0.88)
- DGCLUSTER: A Neural Framework for Attributed Graph Clustering via Modularity Maximization (AAAI 2024 · similarity 0.88)
- Scalable Attribute-Missing Graph Clustering via Neighborhood Differentiation (ICML 2025 · 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.