Research map / Probabilistic and scientific modeling / Variational and generative sampling
On the properties of variational approximations of Gibbs posteriors
Pierre Alquier, James Ridgway, Nicolas Chopin
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 full covariance instead of mean field family (ablation) | VB 0-1 loss · Pima | misclassification rate (%): 31.0 -> 21.3 | Improved |
| use full covariance instead of mean field family (ablation) | VB 0-1 loss · Indian | misclassification rate (%): 48.0 -> 25.5 | Improved |
| use full covariance instead of mean field family (ablation) | VB 0-1 loss · Breast | misclassification rate (%): 35.1 -> 1.1 | Improved |
| use full covariance instead of mean field family (ablation) | VB 0-1 loss · Glass | misclassification rate (%): 34.6 -> 19.6 | Improved |
| use full covariance instead of mean field family (ablation) | VB 0-1 loss · Credit | misclassification rate (%): 32.0 -> 33.6 | Hurt |
| use full covariance instead of mean field family (ablation) | VB 0-1 loss · DNA | misclassification rate (%): 23.6 -> 23.6 | No clear effect |
| replace SMC sampling with full covariance VB | Gibbs posterior · Pima | misclassification rate (%): 22.3 -> 21.3 | Improved |
| replace SMC sampling with full covariance VB | Gibbs posterior · Glass | misclassification rate (%): 23.3 -> 19.6 | Hurt |
| replace 0-1 loss with convex hinge loss in VB | VB Gibbs posterior · DNA | misclassification rate (%): 23.6 -> 4.2 | Improved |
| replace 0-1 loss with convex hinge loss in VB | VB Gibbs posterior · Breast | misclassification rate (%): 1.1 -> 0.5 | Improved |
| replace 0-1 loss with convex hinge loss in VB | VB Gibbs posterior · SPECTF | misclassification rate (%): 06.9 -> 19.2 | Hurt |
| replace 0-1 loss with convex hinge loss in VB | VB Gibbs posterior · Credit | misclassification rate (%): 32.0 -> 27.2 | Improved |
Closest papers by meaning
- Asymptotic Consistency of $\alpha$-{R}ényi-Approximate Posteriors (JMLR 2020 · similarity 0.90)
- Excess Risk Bounds for the Bayes Risk using Variational Inference in Latent Gaussian Models (NeurIPS 2017 · similarity 0.90)
- An Optimization-centric View on Bayes' Rule: Reviewing and Generalizing Variational Inference (JMLR 2022 · similarity 0.89)
- Dissecting Non-Vacuous Generalization Bounds based on the Mean-Field Approximation (ICML 2020 · similarity 0.88)
- Theoretical Guarantees for Variational Inference with Fixed-Variance Mixture of Gaussians (ICML 2024 · similarity 0.87)
- Variational Inference with Mixtures of Isotropic Gaussians (NeurIPS 2025 · similarity 0.87)
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.
