Research map / Learning theory and algorithms / Clustering
Hierarchical clustering with dot products recovers hidden tree structure
Annie Gray, Alexander Modell, Patrick Rubin-Delanchy, Nick Whiteley
Read the paperCheck your idea against 137,217 papers
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 dot product merging instead of UPGMA with Euclidean distance | agglomerative clustering · 20 Newsgroups (raw) | mean Kendall tau_b: 0.23 -> 0.26 | Improved |
| Use dot product merging instead of Ward | agglomerative clustering · Zebrafish (raw) | mean Kendall tau_b: 0.30 -> 0.34 | Improved |
| Use dot product merging instead of UPGMA with cosine distance | agglomerative clustering · Reviews (raw) | mean Kendall tau_b: 0.12 -> 0.15 | Improved |
| Use dot product merging instead of UPGMA with cosine distance | agglomerative clustering · Simulated | mean Kendall tau_b: 0.81 -> 0.86 | Improved |
| Use dot product merging instead of UPGMA with Euclidean distance, Ward or HDBSCAN | agglomerative clustering · Simulated | mean Kendall tau_b: 0.52 -> 0.86 | Improved |
| Use dot product merging instead of UPGMA with cosine distance | agglomerative clustering · S&P 500 (PC scores) | mean Kendall tau_b: 0.42 -> 0.36 | Hurt |
| Use dot product merging instead of Ward | agglomerative clustering · S&P 500 (raw) | mean Kendall tau_b: 0.35 -> 0.34 | No clear effect |
| Use PCA scores instead of raw data vectors as input (ablation) | dot product clustering · 20 Newsgroups | mean Kendall tau_b: 0.26 -> 0.24 | No clear effect |
| Use PCA scores instead of raw data vectors as input (ablation) | UPGMA with Euclidean distance · 20 Newsgroups | mean Kendall tau_b: 0.23 -> 0.038 | Hurt |
| Use PCA scores instead of raw data vectors as input (ablation) | HDBSCAN · S&P 500 | mean Kendall tau_b: 0.14 -> 0.33 | Improved |
| Use PCA-based affinity estimate instead of raw dot product, p growing with n (ablation) | dot product affinity estimation · simulation | maximum affinity estimation error | No clear effect |
| Use PCA-based affinity estimate instead of raw dot product, n fixed at 10 (ablation) | dot product affinity estimation · simulation | maximum affinity estimation error | Hurt |
Closest papers by meaning
- Ultrametric Cluster Hierarchies: I Want ‘em All! (NeurIPS 2025 · similarity 0.89)
- Classification Fields: Arbitrarily Fine Recursive Hierarchical Clustering From Few Examples (NeurIPS 2026 · similarity 0.89)
- From Logits to Hierarchies: Hierarchical Clustering made Simple (ICML 2025 · similarity 0.88)
- Expected Probabilistic Hierarchies (NeurIPS 2024 · similarity 0.87)
- Learning-Augmented Hierarchical Clustering (ICML 2025 · similarity 0.87)
- On the cohesion and separability of average-link for hierarchical agglomerative clustering (NeurIPS 2024 · similarity 0.86)
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.
