AI Paper Atlas

Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes

Model Selection in Bayesian Neural Networks via Horseshoe Priors

Soumya Ghosh, Jiayu Yao, Finale Doshi-Velez

JMLR 2019 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: Approximate inference in Bayesian neural networks · 66 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 horseshoe prior over node pre-activations instead of Gaussian weight priorBayesian neural network · 1-D cubic regression (50/100/1000 units)robustness to under-fittingImproved
use non-centered parameterization for the horseshoe prior (ablation)HS-BNN · 2-D synthetic classification, 15-unit layerstructure recovery (units pruned): prunes away all but two unitsImproved
use horseshoe prior instead of Gaussian priorBNN · MNISTtest errorImproved
use horseshoe prior instead of scale-mixture weight priorSM-BNN · MNISTtest errorNo clear effect
use horseshoe node prior instead of scale-mixture weight priorSM-BNN · MNISTnode sparsityImproved
use structured matrix-variate approximation (VMG) instead of factorized HS-BNNHS-BNN · MNISTtest errorImproved
use horseshoe prior instead of Gaussian priorBNN · gesture recognition (supplement)error rateImproved
use HS-BNN instead of VMGVMG · UCI EnergyRMSE: 0.54 -> 1.99Hurt
use HS-BNN instead of VMGVMG · UCI YachtRMSE: 0.70 -> 1.58Hurt
use HS-BNN instead of VMGVMG · UCI Navaltest log-likelihood: +2.46 -> +5.52Improved
use HS-BNN instead of VMGVMG · UCI BostonRMSE: 2.72 -> 3.32Hurt
use HS-BNN instead of VMGVMG · UCI Power PlantRMSE: 4.04 -> 4.03No clear effect

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