Skip to main navigation Skip to search Skip to main content

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

  • Shuyi Zhang
  • , Chao Pan
  • , Liyan Song
  • , Xiaoyu Wu
  • , Zheng Hu
  • , Ke Pei
  • , Peter Tino
  • , Xin Yao*
  • *Corresponding author for this work
  • Southern University of Science and Technology
  • University of Birmingham
  • Huawei Technologies Co., Ltd.

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

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.

Original languageEnglish
Title of host publicationMachine Learning and Knowledge Discovery in Databases. Research Track - European Conference, ECML PKDD 2021, Proceedings
EditorsNuria Oliver, Fernando Pérez-Cruz, Stefan Kramer, Jesse Read, Jose A. Lozano
PublisherSpringer Science and Business Media Deutschland GmbH
Pages795-810
Number of pages16
ISBN (Print)9783030865221
DOIs
StatePublished - 2021
Externally publishedYes
Event21st Joint European Conference on Machine Learning and Knowledge Discovery in Databases, ECML PKDD 2021 - Virtual, Online
Duration: 13 Sep 202117 Sep 2021

Publication series

NameLecture Notes in Computer Science
Volume12977 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference21st Joint European Conference on Machine Learning and Knowledge Discovery in Databases, ECML PKDD 2021
CityVirtual, Online
Period13/09/2117/09/21

Fingerprint

Dive into the research topics of 'Label-Assisted Memory Autoencoder for Unsupervised Out-of-Distribution Detection'. Together they form a unique fingerprint.

Cite this