Research map / Optimization / Optimizers
Active Slices for Sliced Stein Discrepancy
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 KE active slices and gradient refinement instead of random-initialized gradient slice search | SKSD-g · Laplace, 100D | null rejection rate: 0.58 -> 0.99 | Helped |
| Use KE active slices and gradient refinement instead of random-initialized gradient slice search | SKSD-g · Multi-t, 100D | null rejection rate: 0.67 -> 0.97 | Helped |
| Use KE active slices and gradient refinement instead of random-initialized gradient slice search | SKSD-g · Diffusion, 100D | null rejection rate: 0.96 -> 1 | Helped |
| Use KE active slices rather than gradient-optimized slices | SKSD-rg · RBM GOF, 0.01 perturbation | test power: 0.45 -> 0.67 | Helped |
| Prune to the top-3 most important r slices rather than retain more directions (ablation) | SKSD-rg active slices · RBM GOF | null rejection rate | Helped |
| Add less-important r slices after pruning (ablation) | SKSD-rg active slices · RBM GOF | null rejection rate | Hurt |
| Add short gradient-refinement runs to KE active slices (ablation) | SKSD-rg · RBM GOF | null rejection rate | No clear effect |
| Add gradient refinement to Ex active slices, causing over-fitting to the training set (ablation) | SKSD-rg · RBM GOF | null rejection rate | Hurt |
| Use GE rather than KE for estimating score difference (ablation) | SKSD-rg · RBM GOF | test power | Hurt |
| Use KE instead of exact score differences with refinement (ablation) | SKSD-g ICA model learning · ICA, 80D | test NLL: 15.17$\pm$0.97 -> 7.88$\pm$0.77 | Helped |
| Use KE instead of exact score differences with refinement (ablation) | SKSD-g ICA model learning · ICA, 100D | test NLL: 21.50$\pm$1.41 -> 6.93$\pm$1.36 | Helped |
| Use KE instead of exact score differences with refinement (ablation) | SKSD-g ICA model learning · ICA, 150D | test NLL: 27.37$\pm$3.04 -> 11.67$\pm$2.46 | Helped |
Closest papers by meaning
- Sliced Kernelized Stein Discrepancy (ICLR 2021 · similarity 0.92)
- Generative Sliced MMD Flows with Riesz Kernels (ICLR 2024 · similarity 0.86)
- The Polynomial Stein Discrepancy for Assessing Moment Convergence (ICML 2025 · similarity 0.86)
- Sliced Wasserstein with Random-Path Projecting Directions (ICML 2024 · similarity 0.86)
- Debiased Distribution Compression (ICML 2024 · similarity 0.86)
- Statistical and Computational Guarantees of Kernel Max-Sliced Wasserstein Distances (ICML 2025 · 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.