An Extended Study of Human-like Behavior under Adversarial Training
- Neural networks have a number of shortcomings. Amongst the severest ones is the sensitivity to distribution shifts which allows models to be easily fooled into wrong predictions by small perturbations to inputs that are often imperceivable to humans and do not have to carry semantic meaning. Adversarial training poses a partial solution to address this issue by training models on worst-caseNeural networks have a number of shortcomings. Amongst the severest ones is the sensitivity to distribution shifts which allows models to be easily fooled into wrong predictions by small perturbations to inputs that are often imperceivable to humans and do not have to carry semantic meaning. Adversarial training poses a partial solution to address this issue by training models on worst-case perturbations. Yet, recent work has also pointed out that the reasoning in neural networks is different from humans. Humans identify objects by shape, while neural nets mainly employ texture cues. Exemplarily, a model trained on photographs will likely fail to generalize to datasets containing sketches. Interestingly, it was also shown that adversarial training seems to favorably increase the shift toward shape bias. In this work, we revisit this observation and provide an extensive analysis of this effect on various architectures, the common L_2-and L_-training, and Transformer-based models. Further, we provide a possible explanation for this phenomenon from a frequency perspective.…
Document Type: | Conference Proceeding |
---|---|
Conference Type: | Konferenzartikel |
Zitierlink: | https://opus.hs-offenburg.de/8240 | Bibliografische Angaben |
Title (English): | An Extended Study of Human-like Behavior under Adversarial Training |
Conference: | IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (17-24 June 2023 : Vancouver, BC, Canada) |
Author: | Paul GavrikovStaff MemberORCiDGND, Janis KeuperStaff MemberORCiDGND, Margret Keuper |
Year of Publication: | 2023 |
Creating Corporation: | Computer Vision Foundation |
First Page: | 2361 |
Last Page: | 2368 |
Parent Title (English): | Proceedings : 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops : CVPRW 2023 |
ISBN: | 979-8-3503-0249-3 (Elektronisch) |
ISBN: | 979-8-3503-0250-9 (Print on Demand) |
DOI: | https://doi.org/10.1109/CVPRW59228.2023.00233 |
URL: | https://openaccess.thecvf.com/content/CVPR2023W/AML/papers/Gavrikov_An_Extended_Study_of_Human-Like_Behavior_Under_Adversarial_Training_CVPRW_2023_paper.pdf |
Language: | English | Inhaltliche Informationen |
Institutes: | Fakultät Elektrotechnik, Medizintechnik und Informatik (EMI) (ab 04/2019) |
Forschung / IMLA - Institute for Machine Learning and Analytics | |
Institutes: | Bibliografie |
Tag: | Deep Leaning | Formale Angaben |
Relevance: | Konferenzbeitrag: h5-Index > 30 |
Open Access: | Open Access |
Grün | |
Licence (German): | Urheberrechtlich geschützt |