Volltext-Downloads (blau) und Frontdoor-Views (grau)

Aliasing and adversarial robust generalization of CNNs

  • Many commonly well-performing convolutional neural network models have shown to be susceptible to input data perturbations, indicating a low model robustness. To reveal model weaknesses, adversarial attacks are specifically optimized to generate small, barely perceivable image perturbations that flip the model prediction. Robustness against attacks can be gained by using adversarial examplesMany commonly well-performing convolutional neural network models have shown to be susceptible to input data perturbations, indicating a low model robustness. To reveal model weaknesses, adversarial attacks are specifically optimized to generate small, barely perceivable image perturbations that flip the model prediction. Robustness against attacks can be gained by using adversarial examples during training, which in most cases reduces the measurable model attackability. Unfortunately, this technique can lead to robust overfitting, which results in non-robust models. In this paper, we analyze adversarially trained, robust models in the context of a specific network operation, the downsampling layer, and provide evidence that robust models have learned to downsample more accurately and suffer significantly less from downsampling artifacts, aka. aliasing, than baseline models. In the case of robust overfitting, we observe a strong increase in aliasing and propose a novel early stopping approach based on the measurement of aliasing.…zeige mehrzeige weniger

Metadaten exportieren

Weitere Dienste

Suche bei Google Scholar

Statistik

frontdoor_oas
Metadaten
Dokumentart:Zeitschriftenartikel, wissenschaftlich
Review-Status:Begutachtet (reviewed)
Zitierlink: https://opus.hs-offenburg.de/6445
Bibliografische Angaben
Titel (Englisch):Aliasing and adversarial robust generalization of CNNs
Verfasserangaben:Julia GrabinskiStaff MemberORCiD, Janis KeuperStaff MemberORCiDGND, Margret KeuperORCiD
Erscheinungsjahr:2022
Datum der Erstveröffentlichung:01.11.2022
Verlag:Springer
Erste Seite:3925
Letzte Seite:3951
Titel des übergeordneten Werkes (Englisch):Machine Learning
Jahrgang (Band):111
Heft (Ausgabe):11
ISSN:1573-0565 (Elektronisch)
ISSN:0885-6125 (Print)
DOI:https://doi.org/10.1007/s10994-022-06222-8
Sprache:Englisch
Inhaltliche Informationen
Fakultäten / Einrichtungen:Fakultät Elektrotechnik, Medizintechnik und Informatik (EMI) (ab 04/2019)
Forschung:IMLA - Institute for Machine Learning and Analytics
Freies Schlagwort / Tag:Adversarial robustness; Aliasing; CNNs; Roubst overfitting
Formale Angaben
Relevanz für "Jahresbericht über Forschungsleistungen":5-fach | Wiss. Zeitschriftenartikel reviewed: AGQ-Positivlisten
Open-Access-Status: Open Access 
 Hybrid 
Lizenz (Deutsch):License LogoCreative Commons - CC BY - Namensnennung 4.0 International