Research map / Probabilistic and scientific modeling / Variational and generative sampling
Variational Particle Approximations
Ardavan Saeedi, Tejas D. Kulkarni, Vikash K. Mansinghka, Samuel J. Gershman
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 |
|---|---|---|---|
| increase particles from 1 to 20 in DPVI (ablation) | DPVI · DPMM synthetic D3 | V-measure: 0.51 -> 0.74 | Improved |
| use DPVI K=20 instead of particle filtering K=20 | DPVI · DPMM synthetic D5 | V-measure: 0.05 -> 0.14 | Improved |
| use DPVI K=20 instead of particle filtering K=20 on well-separated clusters | DPVI · DPMM synthetic D1 | V-measure: 0.97 -> 0.99 | No clear effect |
| use DPVI K=100 instead of stratified particle filtering K=1000 | DPVI · Spike sorting | held-out log-likelihood: -3.2965×10^5 -> -1.3888×10^5 | Improved |
| increase DPVI particles from 10 to 100 (ablation) | DPVI · Spike sorting | held-out log-likelihood: -3.2474×10^5 -> -1.3888×10^5 | Improved |
| use DPVI K=20 instead of Gibbs sampling | DPVI · IRM animals | predictive log-likelihood: -374.986 -> -370.674 | Improved |
| increase DPVI particles from 1 to 20 (ablation) | DPVI · IRM animals | predictive log-likelihood: -418.498 -> -370.674 | Improved |
| increase DPVI particles from 10 to 20 (ablation) | DPVI · IRM Alyawarra | predictive log-likelihood: -8.450×10^3 -> -8.450×10^3 | No clear effect |
| use DPVI instead of particle filtering in runtime at K=100 | DPVI · iHMM Alice in Wonderland runtime | seconds: 211.66s -> 1685s | Hurt |
| use DPVI instead of particle filtering in runtime at K=100 | DPVI · DPMM spike sorting runtime | seconds: 454.2s -> 313.8s | Improved |
| use DPVI with 1 or more particles instead of mean-field | DPVI · Ising lattice (beta=100) | lower bound on partition function | Improved |
| add more than 3 particles to DPVI (ablation) | DPVI · Ising lattice (high coupling) | lower bound on partition function | No clear effect |
Closest papers by meaning
- Particle Semi-Implicit Variational Inference (NeurIPS 2024 · similarity 0.88)
- Theoretical Guarantees for Variational Inference with Fixed-Variance Mixture of Gaussians (ICML 2024 · similarity 0.87)
- EigenVI: score-based variational inference with orthogonal function expansions (NeurIPS 2024 · similarity 0.87)
- Efficient Mixture Learning in Black-Box Variational Inference (ICML 2024 · similarity 0.87)
- To discretize continually: Mean shift interacting particle systems for Bayesian inference (NeurIPS 2026 · similarity 0.86)
- Variational Inference with Mixtures of Isotropic Gaussians (NeurIPS 2025 · similarity 0.86)
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.
