Pith. sign in

REVIEW 3 cited by

Being Bayesian, Even Just a Bit, Fixes Overconfidence in ReLU Networks

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2002.10118 v2 pith:4HQZCYE7 submitted 2020-02-24 stat.ML cs.LG

classification stat.MLcs.LG
keywords bayesianrelunetworksapproximateapproximationapproximationsbeencalibrated
verification ladder T0 review T1 audit T2 compute T3 formal
0 comments
read the original abstract

The point estimates of ReLU classification networks---arguably the most widely used neural network architecture---have been shown to yield arbitrarily high confidence far away from the training data. This architecture, in conjunction with a maximum a posteriori estimation scheme, is thus not calibrated nor robust. Approximate Bayesian inference has been empirically demonstrated to improve predictive uncertainty in neural networks, although the theoretical analysis of such Bayesian approximations is limited. We theoretically analyze approximate Gaussian distributions on the weights of ReLU networks and show that they fix the overconfidence problem. Furthermore, we show that even a simplistic, thus cheap, Bayesian approximation, also fixes these issues. This indicates that a sufficient condition for a calibrated uncertainty on a ReLU network is "to be a bit Bayesian". These theoretical results validate the usage of last-layer Bayesian approximation and motivate a range of a fidelity-cost trade-off. We further validate these findings empirically via various standard experiments using common deep ReLU networks and Laplace approximations.

Discussion (0). Sign in to comment.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. Covariance Last-Layer Ensembles: Function-Space Diversity for Efficient Uncertainty Quantification

    cs.LG 2026-07 conditional novelty 6.0 of 10

    Cov-LLE decorrelates member activations in function space, recovering substantial deep-ensemble diversity and calibration at 1× backbone cost, and a direction score repairs OC near-OOD detection.

  2. AGNFormer I: Reconstruction of AGN spectra using a probabilistic transformer model

    astro-ph.GA 2026-07 conditional novelty 6.0 of 10

    An uncertainty-aware transformer reconstructs masked AGN broad lines and spectral halves with 4-16% flux errors and beats eleven purpose-built Lyα-reconstruction algorithms on a blind benchmark.

  3. laplax -- Laplace Approximations with JAX

    cs.LG 2025-07 conditional novelty 6.0 of 10

    The paper presents laplax, a modular JAX library for Laplace approximations that supports multiple curvature estimates, uncertainty pushforwards, calibration, and evaluation routines.

Pith tools