DeepHyperion: Exploring the Feature Space of Deep Learning-based Systems through Illumination Search

Alessio Gambi, Paolo Tonella, Vincenzo Riccio, Tahereh Zohdinasab

Publikation: Beitrag in Buch/Bericht/KonferenzbandKonferenzbeitragBegutachtung

Abstract

In this extended abstract, we summarize our contributions to automated testing of Deep Learning-based systems published at the ACM SIGSOFT International Symposium on Software Testing and Analysis (ISSTA) in 2021 [Zo21a] and just accepted by the ACM Transactions on Software Engineering and Methodology (TOSEM) in 2022 [Zo22]. Deep Learning-based systems (DL Systems) find applications in safety-critical application domains and thus must be thoroughly tested. Existing DL system testing approaches can generate complex and fault-finding inputs but do not characterize them in a way that enables human interpretation and do not always consider test diversity. Our work addresses these challenges and can find effective and diverse test cases.

OriginalspracheEnglisch
TitelSoftware Engineering 2023 - Fachtagung des GI-Fachbereichs Softwaretechnik
Redakteure/-innenGregor Engels, Regina Hebig, Matthias Tichy
Herausgeber (Verlag)Gesellschaft für Informatik e.V.
Seiten131-132
Seitenumfang2
BandP-332
ISBN (elektronisch)9783885797265
PublikationsstatusVeröffentlicht - 24 Feb. 2023

Publikationsreihe

NameLecture Notes in Informatics (LNI), Proceedings - Series of the Gesellschaft fur Informatik (GI)
BandP-332
ISSN (Print)1617-5468

Fingerprint

Untersuchen Sie die Forschungsthemen von „DeepHyperion: Exploring the Feature Space of Deep Learning-based Systems through Illumination Search“. Zusammen bilden sie einen einzigartigen Fingerprint.

Dieses zitieren