Research map / Learning theory and algorithms / Clustering
Measuring the Validity of Clustering Validation Datasets
Hyeon Jeon, Michael J Marie Aupetit, DongHwa Shin, Aeri Cho, Seokhyeon Park, Jinwook Seo
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 |
|---|---|---|---|
| add data-cardinality invariance protocol (T1) alone (ablation) | CH/DI · synthetic Gaussian datasets, data-cardinality test | SMAPE error: about 70% error reduction | Improved |
| add data-cardinality invariance protocol (T1) alone (ablation) | CH/DI · synthetic Gaussian datasets, dimensionality test | SMAPE error: about 30% error reduction | Improved |
| add range invariance protocol (T4) alone (ablation) | IVM · synthetic datasets, data-cardinality test | SMAPE error: about 40% error reduction | Improved |
| add range invariance protocol (T4) alone (ablation) | IVM · synthetic datasets, dimensionality test | SMAPE error: about 55% error reduction | Improved |
| add shift invariance protocol (T2) alone (ablation) | IVM · synthetic datasets | SMAPE error | No clear effect |
| combine shift invariance with data-cardinality or range protocols (ablation) | IVM · synthetic datasets | SMAPE error | Improved |
| apply all possible protocols to SC (only shift invariance) (ablation) | SC · synthetic datasets, dimensionality | SMAPE error | Hurt |
| adjust IVM into IVM_A (average improvement) | IVMs · 96 benchmark datasets across-dataset | Spearman correlation: about 17% (DB) to 81% (DI), average 48% | Improved |
| adjust DB to DB_A | DB · 96 datasets, ami GT | Spearman correlation: .7091 -> .8315 | Improved |
| adjust CH to CH_A | CH · 96 datasets, ami GT | Spearman correlation: .5923 -> .8714 | Improved |
| adjust SC to SC_A | SC · within-dataset noisy labels | Spearman correlation vs NR: .530 -> .878 | Improved |
| adjust CH to CH_A | CH · within-dataset noisy labels | Spearman correlation vs NR: .876 -> .879 | No clear effect |
Closest papers by meaning
- Systematic Analysis of Cluster Similarity Indices: How to Validate Validation Measures (ICML 2021 · similarity 0.88)
- Absolute Cluster Validity (TPAMI 2019 · similarity 0.87)
- Weighted Conformal Clustering (NeurIPS 2026 · similarity 0.85)
- The State of Data Curation at NeurIPS: An Assessment of Dataset Development Practices in the Datasets and Benchmarks Track (NeurIPS 2024 · similarity 0.83)
- Evaluating multiple models using labeled and unlabeled data (NeurIPS 2025 · similarity 0.83)
- You Can Trust Your Clustering Model: A Parameter-free Self-Boosting Plug-in for Deep Clustering (NeurIPS 2025 · similarity 0.83)
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.
