Label-Assisted Memory Autoencoder for Unsupervised Out-of-Distribution Detection

Zhang，Shuyi1,2,3; Pan，Chao1,2; Song，Liyan1,2; Wu，Xiaoyu4; Hu，Zheng4; Pei，Ke5; Tino，Peter3; Yao，Xin1,2,3

10.1007/978-3-030-86523-8_48

2021

0302-9743

1611-3349

LECT NOTES ARTIF INT

12977

795-810

Out-of-Distribution (OoD) detectors based on AutoEncoder (AE) rely on an underlying assumption that an AE network cannot reconstruct OoD data as good as in-distribution (ID) data when it is constructed based on ID data only. However, this assumption may be violated in practice, resulting in a degradation in detection performance. Therefore, alleviating the factors violating this assumption can potentially improve the robustness of OoD performance. Our empirical studies also show that image complexity can be another factor hindering detection performance for AE-based detectors. To cater for these issues, we propose two OoD detectors LAMAE and LAMAE+. Both can be trained without the availability of any OoD-related data. The key idea is to regularize the AE network architecture with a classifier and a label-assisted memory to confine the reconstruction of OoD data while retaining the reconstruction ability for ID data. We also adjust the reconstruction error by taking image complexity into consideration. Experimental studies show that the proposed OoD detectors can perform well on a wider range of OoD scenarios.

第一 ; 通讯

英语

CPCI-S ; EI

Computer Science ; Imaging Science & Photographic Technology

Computer Science, Artificial Intelligence ; Computer Science, Information Systems ; Imaging Science & Photographic Technology

WOS:000713413200048

20213910949555

Complex networks ; Network architecture

Computer Systems and Equipment:722

2-s2.0-85115709907

Scopus

1.Research Institute of Trustworthy Autonomous Systems,Southern University of Science and Technology (SUSTech),Shenzhen,China
2.Guangdong Provincial Key Laboratory of Brain-inspired Intelligent Computation,Department of Computer Science and Engineering,Southern University of Science and Technology (SUSTech),Shenzhen,China
3.CERCIA,School of Computer Science,University of Birmingham,Birmingham,United Kingdom
4.RAMS Reliability Technology Laboratory,Huawei Technology Co. Ltd.,Shenzhen,China
5.TTE-DE RAMS Laboratory,Huawei Technology Co. Ltd.,Munich,Germany

Zhang，Shuyi,Pan，Chao,Song，Liyan,et al. Label-Assisted Memory Autoencoder for Unsupervised Out-of-Distribution Detection[C],2021:795-810.