Research map / Learning theory and algorithms / Clustering
Kernel Clustering: Density Biases and Solutions
Dmitrii Marin, Meng Tang, Ismail Ben Ayed, Yuri Y. Boykov
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 adaptive KNN kernel instead of fixed-bandwidth Gaussian kernel | kernel K-means · LabelMe 4 categories | NMI: 0.38 -> 0.90 | Improved |
| use adaptive KNN kernel instead of basic K-means | kernel K-means · LabelMe 4 categories | NMI: 0.89 -> 0.90 | Improved |
| replace Gaussian with KNN adaptive kernel (average association, no MRF) | average association · GrabCut interactive segmentation | average error (%): 20.4 -> 12.2 | Improved |
| replace Gaussian with KNN adaptive kernel (normalized cut, no MRF) | normalized cut · GrabCut interactive segmentation | average error (%): 17.6 -> 12.4 | Improved |
| replace Gaussian with KNN adaptive kernel (AA, contrast-sensitive MRF) | average association · GrabCut interactive segmentation | average error (%): 9.7 -> 7.1 | Improved |
| add Euclidean length MRF regularization (ablation) | Gaussian AA · GrabCut interactive segmentation | average error (%): 20.4 -> 15.1 | Improved |
| add contrast-sensitive MRF regularization (ablation) | Gaussian AA · GrabCut interactive segmentation | average error (%): 20.4 -> 9.7 | Improved |
| add contrast-sensitive MRF regularization (ablation) | KNN AA · GrabCut interactive segmentation | average error (%): 12.2 -> 7.1 | Improved |
| add Euclidean length MRF regularization (ablation) | KNN NC · GrabCut interactive segmentation | average error (%): 12.4 -> 11.0 | Improved |
| use normalized cut instead of average association with Gaussian kernel (no MRF) | Gaussian kernel · GrabCut interactive segmentation | average error (%): 20.4 -> 17.6 | Improved |
| use normalized cut instead of average association with Gaussian kernel (contrast-sensitive MRF) | Gaussian kernel · GrabCut interactive segmentation | average error (%): 9.7 -> 13.8 | Hurt |
| use normalized cut instead of average association with KNN kernel (contrast-sensitive MRF) | KNN kernel · GrabCut interactive segmentation | average error (%): 7.1 -> 7.8 | Hurt |
Closest papers by meaning
- Implicit Annealing in Kernel Spaces: A Strongly Consistent Clustering Approach (TPAMI 2022 · similarity 0.91)
- Stability and Generalization of Kernel Clustering: from Single Kernel to Multiple Kernel (NeurIPS 2022 · similarity 0.90)
- Adaptive Clustering Using Kernel Density Estimators (JMLR 2023 · similarity 0.90)
- Scalable Multiple Kernel Clustering: Learning Clustering Structure from Expectation (ICML 2024 · similarity 0.89)
- Explaining Kernel Clustering via Decision Trees (ICLR 2024 · similarity 0.87)
- COKE: Core Kernel for More Efficient Approximation of Kernel Weights in Multiple Kernel Clustering (ICML 2025 · similarity 0.87)
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.
