Out-of-Distribution Detection using Multiple Semantic Label Representations

TL;DR

This paper introduces a novel out-of-distribution detection method for neural networks that leverages multiple semantic label representations, improving detection of OOD, misclassified, and adversarial examples across vision and speech tasks.

Contribution

The work proposes using multiple semantic dense representations as label targets, enhancing OOD detection and robustness over traditional sparse label methods.

Findings

Our method outperforms previous approaches in OOD detection.

It effectively detects misclassified and adversarial examples.

The approach is efficient across vision and speech tasks.

Abstract

Deep Neural Networks are powerful models that attained remarkable results on a variety of tasks. These models are shown to be extremely efficient when training and test data are drawn from the same distribution. However, it is not clear how a network will act when it is fed with an out-of-distribution example. In this work, we consider the problem of out-of-distribution detection in neural networks. We propose to use multiple semantic dense representations instead of sparse representation as the target label. Specifically, we propose to use several word representations obtained from different corpora or architectures as target labels. We evaluated the proposed model on computer vision, and speech commands detection tasks and compared it to previous methods. Results suggest that our method compares favorably with previous work. Besides, we present the efficiency of our approach for detecting wrongly classified and adversarial examples.