On the Importance of Regularisation & Auxiliary Information in OOD Detection

TL;DR

This paper introduces two novel objectives that enhance neural networks' ability to detect out-of-distribution samples, reducing overconfidence and improving robustness through regularisation and auxiliary information.

Contribution

The work proposes new methods that outperform existing approaches in OOD detection and highlight the significance of regularisation and auxiliary data in this task.

Findings

Our methods outperform baseline and most existing approaches.

The approach maintains competitive performance under common corruptions.

Regularisation and auxiliary information are crucial for effective OOD detection.

Abstract

Neural networks are often utilised in critical domain applications (e.g. self-driving cars, financial markets, and aerospace engineering), even though they exhibit overconfident predictions for ambiguous inputs. This deficiency demonstrates a fundamental flaw indicating that neural networks often overfit on spurious correlations. To address this problem in this work we present two novel objectives that improve the ability of a network to detect out-of-distribution samples and therefore avoid overconfident predictions for ambiguous inputs. We empirically demonstrate that our methods outperform the baseline and perform better than the majority of existing approaches while still maintaining a competitive performance against the rest. Additionally, we empirically demonstrate the robustness of our approach against common corruptions and demonstrate the importance of regularisation and auxiliary information in out-of-distribution detection.