Energy-based Out-of-distribution Detection

TL;DR

This paper introduces an energy-based framework for out-of-distribution detection that outperforms traditional softmax confidence scores by better distinguishing in- and out-of-distribution samples, improving safety in machine learning deployment.

Contribution

It proposes a unified energy score approach for OOD detection that is theoretically aligned with input density and adaptable to pre-trained classifiers and training procedures.

Findings

Energy scores outperform softmax confidence in OOD detection.

Using energy reduces false positive rate by 18.03% on CIFAR-10.

Energy-based training surpasses existing state-of-the-art methods.

Abstract

Determining whether inputs are out-of-distribution (OOD) is an essential building block for safely deploying machine learning models in the open world. However, previous methods relying on the softmax confidence score suffer from overconfident posterior distributions for OOD data. We propose a unified framework for OOD detection that uses an energy score. We show that energy scores better distinguish in- and out-of-distribution samples than the traditional approach using the softmax scores. Unlike softmax confidence scores, energy scores are theoretically aligned with the probability density of the inputs and are less susceptible to the overconfidence issue. Within this framework, energy can be flexibly used as a scoring function for any pre-trained neural classifier as well as a trainable cost function to shape the energy surface explicitly for OOD detection. On a CIFAR-10 pre-trained WideResNet, using the energy score reduces the average FPR (at TPR 95%) by 18.03% compared to the softmax confidence score. With energy-based training, our method outperforms the state-of-the-art on common benchmarks.