Research map / Probabilistic and scientific modeling / Variational and generative sampling
A Framework for Improving the Reliability of Black-box Variational Inference
Manushi Welandawe, Michael Riis Andersen, Aki Vehtari, Jonathan H. Huggins
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 |
|---|---|---|---|
| use RABVI with termination rule instead of fixed-learning-rate and decay-schedule optimizers | Adam/RMSProp/SGD/ADVI baselines · Gaussian targets | sqrt symmetrized KL | Improved |
| add automated termination rule on top of Adam with exponential decay or cosine schedule | Adam-ED, Adam-C · Gaussian targets | sqrt symmetrized KL | No clear effect |
| use iterate averaging (avgAdam/avgRMSProp) vs other optimizers NGD and SQN inside RABVI (ablation) | RABVI · Gaussian targets | symmetrized KL improvement | Mixed |
| switch from mean-field Gaussian to full-rank Gaussian family (ablation) | RABVI · posteriordb (18 models) | relative mean and std error | Improved |
| switch to real-NVP normalizing flow family (with FASO) (ablation) | FASO · posteriordb (18 models) | relative mean and std error | Mixed |
| compare RABVI to dynamic HMC at equal likelihood evaluations | RABVI · posteriordb (18 models) | relative mean error | Mixed |
| vary initial learning rate, window size, thresholds tau and eps0 (ablation) | RABVI · Gaussian targets, dogs, arK | accuracy and runtime | No clear effect |
| use extreme tuning values (learning rate 0.01 or M at most 5 Monte Carlo samples) (ablation) | RABVI · Gaussian targets, dogs, arK | runtime | Hurt |
Closest papers by meaning
- EigenVI: score-based variational inference with orthogonal function expansions (NeurIPS 2024 · similarity 0.91)
- Batch and match: black-box variational inference with a score-based divergence (ICML 2024 · similarity 0.91)
- Nearly Dimension-Independent Convergence of Mean-Field Black-Box Variational Inference (NeurIPS 2025 · similarity 0.90)
- Efficient Mixture Learning in Black-Box Variational Inference (ICML 2024 · similarity 0.90)
- Generalized Variational Inference via Optimal Transport (AAAI 2024 · similarity 0.90)
- Least squares variational inference (NeurIPS 2025 · 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.
