AI Paper Atlas

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

Feature Preserving Shrinkage on Bayesian Neural Networks Via the R2D2 Prior

Tsai Hor Chan, Ding Zhang, Guosheng Yin, Lequan Yu

TPAMI 2025 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: Approximate inference in Bayesian neural networks · 1 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 variational Gibbs inference instead of SGMCMC with the R2D2 prior (ablation)R2D2-Net · CIFAR-10, AlexNetaccuracy: 55.37 -> 65.10Improved
use variational Gibbs inference instead of SGMCMC with the R2D2 prior (ablation)R2D2-Net · CIFAR-100, AlexNetaccuracy: 20.18 -> 36.12Improved
use variational Gibbs inference instead of SGLD with the R2D2 prior (ablation)R2D2-Net · Simulation S1 (L=3)MSE: 20.61 -> 8.81Improved
use R2D2 prior with SVGI instead of Gaussian prior with SVIBNN · CIFAR-10, AlexNetaccuracy: 60.28 -> 65.10Improved
use R2D2 prior instead of horseshoe priorBNN · CIFAR-100, AlexNetaccuracy: 33.27 -> 36.12Improved
use R2D2 prior instead of horseshoe priorBNN · CIFAR-10, AlexNetaccuracy: 65.01 -> 65.10No clear effect
use R2D2 prior instead of horseshoe prior in OOD detectionBNN · CIFAR-10 vs Fashion MNISTAUROC: 80.76 -> 92.85Improved
use the smaller LeNet architecture instead of AlexNet for R2D2-Net (ablation)R2D2-Net · CIFAR-10accuracy vs frequentist: 62.14 vs 62.24 frequentistNo clear effect
increase MLP depth from L=0 to L=3 for R2D2-Net (ablation)R2D2-Net · Simulation S1MSE: 414.36 -> 8.81Improved
vary hyperparameters a_pi, b and rho_0 of the R2D2 prior (ablation)R2D2-Net · Simulation S1-S3 (L=3)test MSENo 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