Research map / Probabilistic and scientific modeling / Variational and generative sampling
Estimating Density Models with Truncation Boundaries using Score Matching
Song Liu, Takafumi Kanamori, Daniel J. Williams
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 rejection-sampling MLE with TruncSM | truncated Gaussian mixture estimation · Gaussian mixture, polygon truncation | computation time: 3.35 seconds -> 0.35 seconds | Improved |
| increase RJ-MLE particles to many (ablation) | RJ-MLE · Gaussian mixture, polygon truncation | estimation accuracy | Mixed |
| use smaller cap L (0.1 vs 10, 100) in capped weight (ablation) | TruncSM · Gaussian N([1,1],I), 1600 samples, growing domain | estimation error | Improved |
| use l2 distance instead of l1 distance for the weight (ablation) | TruncSM weight function · weight comparison R=0.2 and R=0.8, dim 2-6 | normalized bound quantity: R=0.2 dim 2: 3.160 -> 3.073; dim 6: 8.685 -> 7.552 | Improved |
| use TruncSM instead of vanilla MLE for over-trimmed OSVM data | truncated normal estimation · synthetic 2D inliers with OSVM trimming | 95% confidence region / hold-out likelihood | Improved |
| use TruncSM instead of MLE as trimming gets more aggressive | 20-dim Gaussian inliers · OSVM outlier proportion nu | hold-out likelihood | Improved |
| use TruncSM instead of MLE | per-class Gaussian on PCA-10 CIFAR-10 with 10% outliers · CIFAR-10 (PCA 10-d) | hold-out likelihood | Mixed |
| use TruncSM or RJ-MLE instead of non-truncated MLE | Gaussian mixture on Chicago homicides 2008 · Chicago Crime | estimated component centers | Improved |
Closest papers by meaning
- Sobolev Space Regularised Pre Density Models (ICML 2024 · similarity 0.86)
- ConjNorm: Tractable Density Estimation for Out-of-Distribution Detection (ICLR 2024 · similarity 0.86)
- Density Ratio Estimation with Doubly Strong Robustness (ICML 2024 · similarity 0.84)
- Statistical-Computational Trade-offs for Density Estimation (NeurIPS 2024 · similarity 0.84)
- SD-KDE: Score-Debiased Kernel Density Estimation (NeurIPS 2025 · similarity 0.84)
- Direct Fisher Score Estimation for Likelihood Maximization (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.
