Research map / Probabilistic and scientific modeling / Variational and generative sampling
Variational Weighting for Kernel Density Ratios
Sangwoong Yoon, Frank C. Park, Gunsu S YUN, Iljung Kim, Yung-Kyun Noh
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 |
|---|---|---|---|
| replace KDE with VWKDE in the K-L estimator | KDE-based density ratio · DAGM defect detection | mAUC: 0.734 -> 0.785 | Improved |
| replace KDE with VWKDE | KDE-based density ratio · Unseen defect detection | mAUC: 0.926 -> 0.967 | Improved |
| replace KDE with VWKDE | KDE-based density ratio · DAGM defect localization | mAP: 0.294 -> 0.369 | Improved |
| replace KDE with VWKDE | KDE-based density ratio · Unseen defect localization | mAP: 0.849 -> 0.903 | Improved |
| use unsupervised VWKDE instead of supervised CNN | VWKDE · DAGM defect detection | mAUC: 0.901 -> 0.785 | Hurt |
| use unsupervised VWKDE instead of supervised CNN | VWKDE · Unseen defect detection | mAUC: 0.809 -> 0.967 | Improved |
| use VWKDE instead of KDE for 1D Gaussian log density ratio | KDE · 1D Gaussians, 1000 samples | bias of LPDR | Improved |
| use VWKDE instead of KDE for 1D Gaussian LPDR | KDE · 1D Gaussians | variance of LPDR | Hurt |
| vary bandwidth h for K-L estimate (ablation) | VWKDE · 1D Gaussians | K-L divergence estimate | No clear effect |
| use model-based VWKDE with Gaussian model on 20D NH Gaussians | VWKDE · 20D non-isotropic heteroscedastic Gaussians | K-L divergence accuracy | Improved |
| use misspecified Gaussian score model on Gaussian mixtures (ablation) | VWKDE · 2D and 20D Gaussian mixtures | K-L divergence accuracy | Improved |
Closest papers by meaning
- SD-KDE: Score-Debiased Kernel Density Estimation (NeurIPS 2025 · similarity 0.89)
- Uniform Convergence Rates for Kernel Density Estimation (ICML 2017 · similarity 0.88)
- DiScoFormer: Plug-In Density and Score Estimation with Transformers (ICML 2026 · similarity 0.86)
- Overcoming Saturation in Density Ratio Estimation by Iterated Regularization (ICML 2024 · similarity 0.86)
- Dynamic Similarity Graph Construction with Kernel Density Estimation (ICML 2025 · similarity 0.86)
- Supervised Kernel Thinning (NeurIPS 2024 · 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.
