AI Paper Atlas

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

JMLR 2016 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Annealing and divergence-based variational objectives · 138 citations (OpenAlex)

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 changedWhereMeasuredEffect
use full covariance instead of mean field family (ablation)VB 0-1 loss · Pimamisclassification rate (%): 31.0 -> 21.3Improved
use full covariance instead of mean field family (ablation)VB 0-1 loss · Indianmisclassification rate (%): 48.0 -> 25.5Improved
use full covariance instead of mean field family (ablation)VB 0-1 loss · Breastmisclassification rate (%): 35.1 -> 1.1Improved
use full covariance instead of mean field family (ablation)VB 0-1 loss · Glassmisclassification rate (%): 34.6 -> 19.6Improved
use full covariance instead of mean field family (ablation)VB 0-1 loss · Creditmisclassification rate (%): 32.0 -> 33.6Hurt
use full covariance instead of mean field family (ablation)VB 0-1 loss · DNAmisclassification rate (%): 23.6 -> 23.6No clear effect
replace SMC sampling with full covariance VBGibbs posterior · Pimamisclassification rate (%): 22.3 -> 21.3Improved
replace SMC sampling with full covariance VBGibbs posterior · Glassmisclassification rate (%): 23.3 -> 19.6Hurt
replace 0-1 loss with convex hinge loss in VBVB Gibbs posterior · DNAmisclassification rate (%): 23.6 -> 4.2Improved
replace 0-1 loss with convex hinge loss in VBVB Gibbs posterior · Breastmisclassification rate (%): 1.1 -> 0.5Improved
replace 0-1 loss with convex hinge loss in VBVB Gibbs posterior · SPECTFmisclassification rate (%): 06.9 -> 19.2Hurt
replace 0-1 loss with convex hinge loss in VBVB Gibbs posterior · Creditmisclassification rate (%): 32.0 -> 27.2Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy