AI Paper Atlas

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

Bayesian Scalar-on-Image Regression with a Spatially Varying Single-layer Neural Network Prior

Ben Wu, Keru Wu, Jian Kang

JMLR 2025 · Probabilistic and scientific modeling › Uncertainty and Gaussian processes · approach: Approximate inference in Bayesian neural networks

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
increase hidden units from 1 (STGP) to 5 (ablation)SV-NN · Synthetic nonlinear classification (train size 200)error rate (%): 47.9 -> 15.4Improved
increase hidden units from 3 to 5 (ablation)SV-NN · Synthetic nonlinear classification (train size 200)error rate (%): 29.2 -> 15.4Improved
increase hidden units from 5 to 7 (ablation)SV-NN · Synthetic nonlinear classification (train size 200)error rate (%): 15.4 -> 16.5Hurt
increase hidden units from 5 to 10 (ablation)SV-NN · Synthetic nonlinear classification (train size 200)error rate (%): 15.4 -> 17.2Hurt
use soft threshold 5 instead of 0 (no selection) (ablation)SV-NN · Synthetic nonlinear classification (train size 200)error rate (%): 38.5 -> 15.5Improved
raise threshold from 5 to 6 (ablation)SV-NN · Synthetic nonlinear classification (train size 200)error rate (%): 15.5 -> 22.1Hurt
raise threshold from 5 to 7 (ablation)SV-NN · Synthetic nonlinear classification (train size 200)error rate (%): 15.5 -> 28.4Hurt
use MSE kernel instead of squared exponential kernel (ablation)SV-NN · Synthetic nonlinear classification (train size 200)error rate (%): 17.9 -> 15.4Improved
use MSE kernel instead of Matern kernel (ablation)SV-NN · Synthetic nonlinear classification (train size 200)error rate (%): 18.6 -> 15.4Improved
use MSE kernel instead of exponential kernel (ablation)SV-NN · Synthetic nonlinear classification (train size 200)error rate (%): 19.2 -> 15.4Improved
use SV-NN instead of CNN with matching structureCNN · Synthetic nonlinear classification (train size 200)error rate (%): 20.4 -> 15.5Improved
use SV-NN instead of BNN with limited samplesBNN · MNIST 4 vs 7 (n=10 per digit)error rate (%): 18.3 -> 4.6Improved

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