Research map / Learning theory and algorithms / Clustering
Near-perfect Clustering Based on Recursive Binary Splitting Using Max-MMD
Sourav Chakrabarty, Anirvan Chakraborty, Shyamal K. De
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 |
|---|---|---|---|
| increase cluster sizes from (30,30) to (100,100) | CURBS-I · Sim K=2 Model 4 | Rand index: 0.492 -> 0.847 | Improved |
| increase cluster sizes from (30,30,30) to (100,100,100) on Model 8 | CURBS-I · Sim K=3 Model 8 | Rand index: 0.827 -> 0.981 | Improved |
| use unbalanced sizes (150,25,25) instead of (100,100,100) on Model 8 | CURBS-I · Sim K=3 Model 8 | Rand index: 0.981 -> 0.592 | Hurt |
| use MMD-based CURBS-II instead of L2PC k-means on scale change | CURBS-II · Sim K=2 (30,30) Model 1(b) | Rand index: 0.53 -> 0.984 | Improved |
| use CURBS-II instead of L2PC on location change Model 1(a) | CURBS-II · Sim K=2 (30,30) Model 1(a) | Rand index: 0.983 -> 0.972 | Hurt |
| use CURBS-II instead of kCFC on Model 7(a) | CURBS-II · Sim K=3 (30,30,30) Model 7(a) | Rand index: 1 -> 0.98 | Hurt |
| use CURBS-II instead of DHP_HA on Model 4 eigenfunction change | CURBS-II · Sim K=2 (30,30) Model 4 | Rand index: 0.5 -> 0.907 | Improved |
| specify K in advance (CURBS-II) instead of estimating it (CURBS-I) | CURBS · Phoneme | Rand index: 0.896 -> 0.922 | Improved |
| use CURBS-II instead of L2PC | CURBS-II · Phoneme | Rand index: 0.918 -> 0.922 | No clear effect |
| vary Laplacian kernel bandwidth h among 0.1, 0.2, 1, 5, 10 instead of median heuristic (ablation) | CURBS · simulated models | Rand index | No clear effect |
| increase sample size n from 30 to 100 for single-population check | CURBS-I SCC · Model N2 | proportion K_hat=1: 95% -> 98% | Improved |
Closest papers by meaning
- A Constant Approximation Algorithm for Sequential Random-Order No-Substitution k-Median Clustering (NeurIPS 2021 · similarity 0.86)
- Uniform Concentration Bounds toward a Unified Framework for Robust Clustering (NeurIPS 2021 · similarity 0.86)
- Efficient Parameter-Free Clustering Using First Neighbor Relations (CVPR 2019 · similarity 0.86)
- Parameterized Approximation Algorithms for Sum of Radii Clustering and Variants (AAAI 2024 · similarity 0.85)
- Beyond Worst-Case Coreset Bounds for $k$-Clustering via Determinantal Sampling (NeurIPS 2026 · similarity 0.85)
- Competitively Consistent Clustering (ICML 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.
