Research map / Optimization / Optimizers
Black-Box Alpha Divergence Minimization
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 |
|---|---|---|---|
| alpha=-1.0 instead of alpha=1.0 (ablation) | BB-α · MNIST | average test error/100: 1.51 -> 1.33 | Helped |
| alpha=-1.0 instead of alpha=1.0 (ablation) | BB-α · MNIST | average test log-likelihood/100: -5.51 -> -4.47 | Helped |
| alpha=10^-6 vs BB-VB (ablation) | BB-α · MNIST | average test error/100: 1.36 -> 1.36 | No clear effect |
| alpha 0.5 instead of 1.0 (ablation) | BB-α · Clean Energy Project | average test error: 1.28 -> 1.08 | Helped |
| alpha 0.5 instead of 1.0 (ablation) | BB-α · Clean Energy Project | average test log-likelihood: -0.93 -> -0.74 | Helped |
| alpha=10^-6 vs BB-VB (ablation) | BB-α · Clean Energy Project | average test error: 1.13 -> 1.14 | No clear effect |
| alpha 1.0 instead of BB-VB (ablation) | BB-α · Probit regression, Madelon | average test log-likelihood: -0.953 -> -0.799 | Helped |
| alpha 1.0 instead of BB-VB (ablation) | BB-α · Probit regression, Ionosphere | average test log-likelihood: -0.333 -> -0.333 | No clear effect |
| alpha 1.0 instead of alpha 10^-6 (ablation) | BB-α · Probit regression | average rank (test log-likelihood): 2.970 -> 1.895 | Helped |
| alpha=10^-6 vs BB-VB (ablation) | BB-α · Probit regression | average rank (test log-likelihood): 2.970 -> 2.845 | No clear effect |
| increase Monte Carlo samples K from 1 to 10 (ablation) | BB-α=1.0 · Boston housing gradient | average bias: 0.2774 -> 0.0077 | Helped |
| alpha 10^-6 instead of 1.0 at K=1 (ablation) | BB-α · Boston housing gradient | average bias: 0.2774 -> 0.0460 | Helped |
Closest papers by meaning
- Batch and match: black-box variational inference with a score-based divergence (ICML 2024 · similarity 0.87)
- Perturbative Black Box Variational Inference (NeurIPS 2017 · similarity 0.86)
- Asymptotics of Alpha-Divergence Variational Inference Algorithms with Exponential Families (NeurIPS 2024 · similarity 0.85)
- Posterior Inference with Diffusion Models for High-dimensional Black-box Optimization (ICML 2025 · similarity 0.83)
- Efficient Mixture Learning in Black-Box Variational Inference (ICML 2024 · similarity 0.82)
- A Black-Box Debiasing Framework for Conditional Sampling (NeurIPS 2025 · similarity 0.82)
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.