Research map / Learning theory and algorithms / Clustering
Efficient and Local Parallel Random Walks
Michael Kapralov, Silvio Lattanzi, Navid Nouri, Jakab Tardos
AI-extracted. The measured effects below were read by AI from the paper's full text (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 adaptive local stitching instead of Uniform Stitching | Random-walk generation from 15 roots · com-DBLP | runtime: $66\pm1$ minutes -> $18\pm1$ minutes | Improved |
| Use adaptive local stitching instead of Uniform Stitching | Random-walk generation from 15 roots · email-Enron | runtime: $15\pm0$ minutes -> $19\pm1$ minutes | Improved |
| Use adaptive local stitching instead of Uniform Stitching | Random-walk generation from 15 roots · ca-GrQc | runtime: $7\pm0$ minutes -> $15\pm1$ minutes | Hurt |
| Use adaptive local stitching instead of Uniform Stitching | Random-walk generation from 15 roots with λ=20 · email-Enron and com-DBLP | completion: Uniform Stitching: --- | Improved |
| Lower B₀ and τ (ablation) | Local Stitching · SNAP graphs, ℓ=16 | runtime: com-DBLP: $23\pm7$ minutes -> $17\pm1$ minutes; com-Youtube: $34\pm6$ minutes -> $23\pm2$ minutes; com-LiveJournal: $76\ | Improved |
| Lower B₀ and τ (ablation) | Local Stitching · SNAP graphs, ℓ=16 | walk failure rate: com-DBLP: 14.6\pm0.6\% -> 38.3\pm0.7\%; com-Youtube: 10.8\pm0.5\% -> 28.1\pm1.7\%; com-LiveJournal: 7.9\pm0.1\% -> 26.2\ | Hurt |
| Increase walk length from ℓ=16 to ℓ=32 (ablation) | Local Stitching · com-DBLP and com-Youtube | runtime: com-DBLP: $23\pm7$ minutes -> $25\pm2$ minutes; com-Youtube: $34\pm6$ minutes -> $45\pm1$ minutes | Mixed |
| Increase walk length from ℓ=16 to ℓ=32 (ablation) | Local Stitching · com-LiveJournal and com-Orkut | runtime: com-LiveJournal: $76\pm11$ minutes -> $115\pm3$ minutes; com-Orkut: $64\pm13$ minutes -> $95\pm1$ minutes | Hurt |
| Increase λ from 10 to 20 (ablation) | Local Stitching · ca-GrQc, email-Enron, and com-DBLP | runtime: ca-GrQc: $15\pm1$ minutes -> $15\pm1$ minutes; email-Enron: $19\pm1$ minutes -> $19\pm1$ minutes; com-DBLP: $18\pm1$ min | No clear effect |
Closest papers by meaning
- Weighted Flow Diffusion for Local Graph Clustering with Node Attributes: an Algorithm and Statistical Guarantees (ICML 2023 · similarity 0.87)
- Capacity Releasing Diffusion for Speed and Locality (ICML 2017 · similarity 0.86)
- Local Graph Clustering with Noisy Labels (ICLR 2024 · similarity 0.86)
- Revisiting Random Walks for Learning on Graphs (ICLR 2025 · similarity 0.86)
- New Parallel and Streaming Algorithms for Directed Densest Subgraph (NeurIPS 2025 · similarity 0.84)
- A Near-Linear Time Approximation Algorithm for Beyond-Worst-Case Graph Clustering (ICML 2024 · similarity 0.84)
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.
