Research map / Label-efficient and robust learning / Recognition and long-tail
Modeling Skewed Class Distributions by Reshaping the Concept Space
Kyle Feuz, Diane Cook
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 |
|---|---|---|---|
| decompose majority classes into sub-classes with EM clustering (ICC EM All) | Naive Bayes · 15 imbalanced datasets | accuracy (# significant wins): 9 better, 0 worse | Improved |
| decompose classes with ICC | J48 decision tree · 15 imbalanced datasets | accuracy (# significant wins): 0 better, 4 worse | No clear effect |
| ICC EM All instead of no imbalance technique | LibLINEAR · 15 imbalanced datasets | accuracy (# significant wins): 2 better, 0 worse | Improved |
| ICC instead of Resample or Balance weighting | Naive Bayes · 15 imbalanced datasets | accuracy (# significant losses): Resample 10 worse, Balance 11 worse, ICC EM All 0 worse | Improved |
| apply ICC to all classes instead of majority classes only (ablation) | ICC · 15 imbalanced datasets | accuracy and F1 (# significant wins) | No clear effect |
| choose number of clusters automatically via EM instead of fixed k=15 (ablation) | ICC All · 15 imbalanced datasets | accuracy (# significant wins): Naive Bayes 9 vs 7 | Improved |
| increase fixed number of sub-classes k from 2 to 15 (ablation) | ICC Fix All · 15 imbalanced datasets | accuracy | Mixed |
| apply ICC to datasets without intrinsic sub-classes (yeast, ecoli) | ICC · yeast, ecoli | accuracy: yeast best F1 0.00 (NB) | No clear effect |
| apply ICC EM/CSK to majority classes only (ablation) | class distribution · 15 datasets | KL-divergence change: e.g. -1.27 on letters-cons | Improved |
| apply ICC EM/CSK to all classes (ablation) | class distribution · 15 datasets | KL-divergence change: e.g. 0.49 on yeast6, -0.85 on letters-cons | Mixed |
Closest papers by meaning
- Balancing the Scales: A Theoretical and Algorithmic Framework for Learning from Imbalanced Data (ICML 2025 · similarity 0.88)
- Balanced Data, Imbalanced Spectra: Unveiling Class Disparities with Spectral Imbalance (ICML 2024 · similarity 0.85)
- Enhancing Class-Imbalanced Learning with Pre-Trained Guidance through Class-Conditional Knowledge Distillation (ICML 2024 · similarity 0.85)
- Class-Imbalanced Graph Learning without Class Rebalancing (ICML 2024 · similarity 0.85)
- Learning under Localized Minority Imbalance (NeurIPS 2026 · similarity 0.85)
- Class-Attribute Priors: Adapting Optimization to Heterogeneity and Fairness Objective (AAAI 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.
