WOOD: Wasserstein-Based Out-of-Distribution Detection

• Activity Number: 58 - Q&P and SPES Student Paper Award
• Type: Topic Contributed
• Date/Time: Sunday, August 7, 2022 : 4:00 PM to 5:50 PM
• Sponsor: Quality and Productivity Section
• Author(s): Yinan Wang* and Wenbo Sun and Judy Jin and Zhenyu Kong and Xiaowei Yue
• Companies: Virginia Tech and University of Michigan Transportation Research Institute and University of Michigan and Virginia Tech and Virginia Tech
• Keywords: OOD detection; Wasserstein distance; machine learning; image classification; cyber security

Abstract

The training and test data for deep-neural-network-based classifiers are usually assumed to be sampled from the same distribution. When part of the test samples are drawn from a different distribution (a.k.a. out-of-distribution (OOD) samples), the trained neural network is prone to be fooled by these OOD samples. Detection of OOD samples has three main challenges: (i) the proposed OOD detection method should be compatible with various architectures of classifiers; (ii) the labels of OOD samples are commonly unavailable; (iii) a score function needs to be defined to effectively identify OOD samples. We propose a Wasserstein-based out-of-distribution detection (WOOD) method, which defines a Wasserstein-distance-based score that evaluates the dissimilarity between a test sample and the distribution of InD samples. An optimization problem is then formulated and solved based on the proposed score function. The statistical learning bound is investigated to guarantee that the loss value achieved by the empirical optimizer approximates the global optimum. The comparison study results demonstrate that the proposed WOOD consistently outperforms other existing OOD detection methods.