Towards neural networks that provably know when they don't know

Abstract

It has recently been shown that ReLU networks produce arbitrarily over-confident predictions far away from the training data. Thus, ReLU networks do not know when they don't know. However, this is a highly important property in safety critical applications. In the context of out-of-distribution detection (OOD) there have been a number of proposals to mitigate this problem but none of them are able to make any mathematical guarantees. In this paper we propose a new approach to OOD which overcomes both problems. Our approach can be used with ReLU networks and provides provably low confidence predictions far away from the training data as well as the first certificates for low confidence predictions in a neighborhood of an out-distribution point. In the experiments we show that state-of-the-art methods fail in this worst-case setting whereas our model can guarantee its performance while retaining state-of-the-art OOD performance.

References in corpus (9)

• Auto-Encoding Variational Bayes
• Stochastic Backpropagation and Approximate Inference in Deep Generative Models
• Provable defenses against adversarial examples via the convex outer adversarial polytope
• Certified Defenses against Adversarial Examples
• Evidential Deep Learning to Quantify Classification Uncertainty
• Early Methods for Detecting Adversarial Images
• Open Category Detection with PAC Guarantees
• Safer Classification by Synthesis
• Better the Devil you Know: An Analysis of Evasion Attacks using Out-of-Distribution Adversarial Examples

Cited by in corpus (3)

• Natural Adversarial Examples
• A Smooth Representation of Belief over SO(3) for Deep Rotation Learning with Uncertainty
• Unsupervised Anomaly Detection From Semantic Similarity Scores