Research map / Optimization / Optimizers
Fast Black-box Variational Inference through Stochastic Trust-Region Optimization
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 TrustVI instead of ADVI | Variational inference on the study models · 183 statistical models and datasets | models where the method found a better median optimal ELBO: 51 models (28%) | Helped |
| Use TrustVI instead of ADVI | Variational inference on the study models · 183 statistical models and datasets | models where the method found a worse median optimal ELBO: 6 models (3%) | Hurt |
| Use TrustVI instead of ADVI | Variational inference on the runtime-comparison models · ADVI runtime comparison | runtime: 136 of 137 models (99%); 69 models (50%) at least 12x faster; 34 models (25%) at least 36x faster | Helped |
| Use TrustVI instead of HFSGVI | Variational inference on the study models · 160 statistical models and datasets | models where the method found a better median optimal ELBO: 51 models (28%) | Helped |
| Use TrustVI instead of HFSGVI | Variational inference on the study models · 160 statistical models and datasets | models where the method found a worse median optimal ELBO: 1 model (0.5%) | Hurt |
| Use TrustVI instead of HFSGVI | Variational inference on runtime-comparison models · HFSGVI runtime comparison | runtime: 61 models; HFSGVI was faster for 54 models | Mixed |
| Use trust-region step assessment and rejection | TrustVI · HFSGVI comparison | numerical overflow failures: 23 of the 183 models | Helped |
| Increase ADVI adaptation iterations | ADVI · ADVI optimization | uniform improvement across settings | No clear effect |
Closest papers by meaning
- Robust, Accurate Stochastic Optimization for Variational Inference (NeurIPS 2020 · similarity 0.92)
- Nearly Dimension-Independent Convergence of Mean-Field Black-Box Variational Inference (NeurIPS 2025 · similarity 0.90)
- Least squares variational inference (NeurIPS 2025 · similarity 0.89)
- Natural Gradient VI: Guarantees for Non-Conjugate Models (NeurIPS 2025 · similarity 0.89)
- Revisiting Unbiased Implicit Variational Inference (ICML 2025 · similarity 0.89)
- EigenVI: score-based variational inference with orthogonal function expansions (NeurIPS 2024 · similarity 0.89)
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.