Research map / Learning theory and algorithms / Clustering
An Efficient and Effective Generic Agglomerative Hierarchical Clustering Approach
Julien Ah-Pine
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 |
|---|---|---|---|
| sparsify similarity matrix to 1% nearest neighbors (ablation) | Ward AHC · Aggregation | ARI: 0.688 -> 0.965 | Improved |
| sparsify similarity matrix to 1% nearest neighbors (ablation) | Group average AHC · Aggregation | ARI: 0.991 -> 1.000 | Improved |
| sparsify with threshold to keep top 1% similarities (ablation) | Ward AHC · Compound | ARI: 0.501 -> 0.906 | Improved |
| sparsify similarity matrix to 1% nearest neighbors (ablation) | W-Median AHC · Aggregation | ARI: 0.780 -> 0.590 | Hurt |
| sparsify similarity matrix to 10% nearest neighbors (ablation) | Median AHC · Aggregation | ARI: 0.996 -> 0.415 | Hurt |
| keep 50% of nearest neighbors (ablation) | All six AHC schemes · Aggregation | ARI: 0.991 -> 0.991 (group average) | No clear effect |
| threshold away 50% of similarities (ablation) | All six AHC schemes · Compound | ARI: 0.811 -> 0.811 (group average) | No clear effect |
| sparsify similarity matrix to k=round(n/10) (ablation) | Group average AHC · Landsat | ARI: 0.321 -> 0.688 | Improved |
| keep k>=round(n/2) nearest neighbors, halving memory and runtime (ablation) | SNK-AHC · Landsat | ARI | No clear effect |
| reduce similarity matrix to 10% of its memory size (ablation) | SNK-AHC · Landsat | relative running time: running time reduced to 10% | Improved |
| threshold 75-90% of similarity values (ablation) | SNK-AHC · Compound | ARI | Mixed |
Closest papers by meaning
- On the cohesion and separability of average-link for hierarchical agglomerative clustering (NeurIPS 2024 · similarity 0.90)
- Ultrametric Cluster Hierarchies: I Want ‘em All! (NeurIPS 2025 · similarity 0.88)
- New Bounds on the Cohesion of Complete-link and Other Linkage Methods for Agglomerative Clustering (ICML 2024 · similarity 0.87)
- From Logits to Hierarchies: Hierarchical Clustering made Simple (ICML 2025 · similarity 0.87)
- Hierarchical Overlapping Clustering on Graphs: Cost Function, Algorithm and Scalability (ICML 2025 · similarity 0.86)
- Efficient Centroid-Linkage 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.
