Adversarial Robustness Via Fisher-Rao Regularization

Adversarial Robustness Via Fisher-Rao Regularization

Adversarial robustness has become a topic of growing interest in machine learning since it was observed that neural networks tend to be brittle. We propose an information-geometric formulation of adversarial defense and introduce Fire, a new Fisher-Rao regularization for the categorical cross-entrop...

Ausführliche Beschreibung

Bibliographische Detailangaben
Veröffentlicht in:IEEE transactions on pattern analysis and machine intelligence. - 1979. - 45(2023), 3 vom: 12. März, Seite 2698-2710
1. Verfasser: Picot, Marine (VerfasserIn)
Weitere Verfasser: Messina, Francisco, Boudiaf, Malik, Labeau, Fabrice, Ayed, Ismail Ben, Piantanida, Pablo
Format: Online-Aufsatz
Sprache:English
Veröffentlicht: 2023
Zugriff auf das übergeordnete Werk:IEEE transactions on pattern analysis and machine intelligence
Schlagworte:Journal Article
LEADER 01000naa a22002652 4500
001 NLM340800755
003 DE-627
005 20231226010305.0
007 cr uuu---uuuuu
008 231226s2023 xx |||||o 00| ||eng c
024 7 |a 10.1109/TPAMI.2022.3174724  |2 doi 
028 5 2 |a pubmed24n1135.xml 
035 |a (DE-627)NLM340800755 
035 |a (NLM)35552150 
040 |a DE-627  |b ger  |c DE-627  |e rakwb 
041 |a eng 
100 1 |a Picot, Marine  |e verfasserin  |4 aut 
245 1 0 |a Adversarial Robustness Via Fisher-Rao Regularization 
264 1 |c 2023 
336 |a Text  |b txt  |2 rdacontent 
337 |a ƒaComputermedien  |b c  |2 rdamedia 
338 |a ƒa Online-Ressource  |b cr  |2 rdacarrier 
500 |a Date Revised 20.10.2023 
500 |a published: Print-Electronic 
500 |a Citation Status PubMed-not-MEDLINE 
520 |a Adversarial robustness has become a topic of growing interest in machine learning since it was observed that neural networks tend to be brittle. We propose an information-geometric formulation of adversarial defense and introduce Fire, a new Fisher-Rao regularization for the categorical cross-entropy loss, which is based on the geodesic distance between the softmax outputs corresponding to natural and perturbed input features. Based on the information-geometric properties of the class of softmax distributions, we derive an explicit characterization of the Fisher-Rao Distance (FRD) for the binary and multiclass cases, and draw some interesting properties as well as connections with standard regularization metrics. Furthermore, we verify on a simple linear and Gaussian model, that all Pareto-optimal points in the accuracy-robustness region can be reached by Fire while other state-of-the-art methods fail. Empirically, we evaluate the performance of various classifiers trained with the proposed loss on standard datasets, showing up to a simultaneous 1% of improvement in terms of clean and robust performances while reducing the training time by 20% over the best-performing methods 
650 4 |a Journal Article 
700 1 |a Messina, Francisco  |e verfasserin  |4 aut 
700 1 |a Boudiaf, Malik  |e verfasserin  |4 aut 
700 1 |a Labeau, Fabrice  |e verfasserin  |4 aut 
700 1 |a Ayed, Ismail Ben  |e verfasserin  |4 aut 
700 1 |a Piantanida, Pablo  |e verfasserin  |4 aut 
773 0 8 |i Enthalten in  |t IEEE transactions on pattern analysis and machine intelligence  |d 1979  |g 45(2023), 3 vom: 12. März, Seite 2698-2710  |w (DE-627)NLM098212257  |x 1939-3539  |7 nnns 
773 1 8 |g volume:45  |g year:2023  |g number:3  |g day:12  |g month:03  |g pages:2698-2710 
856 4 0 |u http://dx.doi.org/10.1109/TPAMI.2022.3174724  |3 Volltext 
912 |a GBV_USEFLAG_A 
912 |a SYSFLAG_A 
912 |a GBV_NLM 
912 |a GBV_ILN_350 
951 |a AR 
952 |d 45  |j 2023  |e 3  |b 12  |c 03  |h 2698-2710