Skip to main navigation Skip to search Skip to main content

Gaussian-Based Runtime Detection of Out-of-distribution Inputs for Neural Networks

  • AUDI AG
  • Technical University of Munich

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

9 Scopus citations

Abstract

In this short paper, we introduce a simple approach for runtime monitoring of deep neural networks and show how to use it for out-of-distribution detection. The approach is based on inferring Gaussian models of some of the neurons and layers. Despite its simplicity, it performs better than recently introduced approaches based on interval abstractions which are traditionally used in verification.

Original languageEnglish
Title of host publicationRuntime Verification - 21st International Conference, RV 2021, Proceedings
EditorsLu Feng, Dana Fisman
PublisherSpringer Science and Business Media Deutschland GmbH
Pages254-264
Number of pages11
ISBN (Print)9783030884932
DOIs
StatePublished - 2021
Event21st International Conference on Runtime Verification, RV 2021 - Virtual, Online
Duration: 11 Oct 202114 Oct 2021

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume12974 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference21st International Conference on Runtime Verification, RV 2021
CityVirtual, Online
Period11/10/2114/10/21

Fingerprint

Dive into the research topics of 'Gaussian-Based Runtime Detection of Out-of-distribution Inputs for Neural Networks'. Together they form a unique fingerprint.

Cite this