Uncertainty Calibration in Bayesian Neural Networks via Distance-Aware Priors

2022/07/17 by Gianluca Detommaso, Detommaso, Gianluca, Alberto Gasparin +5 · 2 citations
Computer Science · Engineering · #Adversarial Robustness in Machine Learning #Artificial Intelligence (cs.AI) #FOS: Computer and information sciences #Fault Detection and Control Systems #Gaussian Processes and Bayesian Inference #Machine Learning (cs.LG) #Machine Learning (stat.ML)

paper · pdf · doi:10.48550/arxiv.2207.08200

openalex publication_date 2022/07/17 · openalex created_date 2022/07/21 · openalex updated_date 2026/07/28

Abstract

As we move away from the data, the predictive uncertainty should increase, since a great variety of explanations are consistent with the little available information. We introduce Distance-Aware Prior (DAP) calibration, a method to correct overconfidence of Bayesian deep learning models outside of the training domain. We define DAPs as prior distributions over the model parameters that depend on the inputs through a measure of their distance from the training set. DAP calibration is agnostic to the posterior inference method, and it can be performed as a post-processing step. We demonstrate its effectiveness against several baselines in a variety of classification and regression problems, including benchmarks designed to test the quality of predictive distributions away from the data.

Cited by

• From Overfitting to Reliability: Introducing the Hierarchical Approximate Bayesian Neural Network
• Where are we with calibration under dataset shift in image classification?

Related

• Improving Uncertainty Calibration via Prior Augmented Data
• Accurate Uncertainties for Deep Learning Using Calibrated Regression
• Simple and Principled Uncertainty Estimation with Deterministic Deep Learning via Distance Awareness
• On Calibrated Model Uncertainty in Deep Learning
• Beyond In-Domain Scenarios: Robust Density-Aware Calibration