Matches in SemOpenAlex for { <https://semopenalex.org/work/W3166379334> ?p ?o ?g. }
Showing items 1 to 94 of
94
with 100 items per page.
- W3166379334 abstract "Adversarial training is one of the most effective approaches to improve model robustness against adversarial examples. However, previous works mainly focus on the overall robustness of the model, and the in-depth analysis on the role of each class involved in adversarial training is still missing. In this paper, we propose to analyze the class-wise robustness in adversarial training. First, we provide a detailed diagnosis of adversarial training on six benchmark datasets, i.e., MNIST, CIFAR-10, CIFAR-100, SVHN, STL-10 and ImageNet. Surprisingly, we find that there are remarkable robustness discrepancies among classes, leading to unbalance/unfair class-wise robustness in the robust models. Furthermore, we keep investigating the relations between classes and find that the unbalanced class-wise robustness is pretty consistent among different attack and defense methods. Moreover, we observe that the stronger attack methods in adversarial learning achieve performance improvement mainly from a more successful attack on the vulnerable classes (i.e., classes with less robustness). Inspired by these interesting findings, we design a simple but effective attack method based on the traditional PGD attack, named Temperature-PGD attack, which proposes to enlarge the robustness disparity among classes with a temperature factor on the confidence distribution of each image. Experiments demonstrate our method can achieve a higher attack rate than the PGD attack. Furthermore, from the defense perspective, we also make some modifications in the training and inference phase to improve the robustness of the most vulnerable class, so as to mitigate the large difference in class-wise robustness. We believe our work can contribute to a more comprehensive understanding of adversarial training as well as rethinking the class-wise properties in robust models." @default.
- W3166379334 created "2021-06-22" @default.
- W3166379334 creator A5004999983 @default.
- W3166379334 creator A5012288572 @default.
- W3166379334 creator A5041727387 @default.
- W3166379334 creator A5064753974 @default.
- W3166379334 creator A5070320567 @default.
- W3166379334 date "2021-05-29" @default.
- W3166379334 modified "2023-10-01" @default.
- W3166379334 title "Analysis and Applications of Class-wise Robustness in Adversarial Training" @default.
- W3166379334 cites W1673923490 @default.
- W3166379334 cites W2108598243 @default.
- W3166379334 cites W2112796928 @default.
- W3166379334 cites W2118858186 @default.
- W3166379334 cites W2194775991 @default.
- W3166379334 cites W2335728318 @default.
- W3166379334 cites W2787708942 @default.
- W3166379334 cites W2885707361 @default.
- W3166379334 cites W2913266441 @default.
- W3166379334 cites W2943646750 @default.
- W3166379334 cites W2945793108 @default.
- W3166379334 cites W2950787360 @default.
- W3166379334 cites W2954978443 @default.
- W3166379334 cites W2963207607 @default.
- W3166379334 cites W2963857521 @default.
- W3166379334 cites W2964253222 @default.
- W3166379334 cites W2966913949 @default.
- W3166379334 cites W2970680991 @default.
- W3166379334 cites W2985987886 @default.
- W3166379334 cites W2995056242 @default.
- W3166379334 cites W2995645057 @default.
- W3166379334 cites W2996564870 @default.
- W3166379334 cites W3006076803 @default.
- W3166379334 cites W3089060762 @default.
- W3166379334 cites W3101119695 @default.
- W3166379334 cites W3103385169 @default.
- W3166379334 cites W3104182862 @default.
- W3166379334 cites W3104620392 @default.
- W3166379334 cites W3118608800 @default.
- W3166379334 cites W3125614726 @default.
- W3166379334 cites W3134501548 @default.
- W3166379334 cites W3137894200 @default.
- W3166379334 doi "https://doi.org/10.48550/arxiv.2105.14240" @default.
- W3166379334 hasPublicationYear "2021" @default.
- W3166379334 type Work @default.
- W3166379334 sameAs 3166379334 @default.
- W3166379334 citedByCount "1" @default.
- W3166379334 countsByYear W31663793342021 @default.
- W3166379334 crossrefType "posted-content" @default.
- W3166379334 hasAuthorship W3166379334A5004999983 @default.
- W3166379334 hasAuthorship W3166379334A5012288572 @default.
- W3166379334 hasAuthorship W3166379334A5041727387 @default.
- W3166379334 hasAuthorship W3166379334A5064753974 @default.
- W3166379334 hasAuthorship W3166379334A5070320567 @default.
- W3166379334 hasBestOaLocation W31663793341 @default.
- W3166379334 hasConcept C104317684 @default.
- W3166379334 hasConcept C119857082 @default.
- W3166379334 hasConcept C154945302 @default.
- W3166379334 hasConcept C185592680 @default.
- W3166379334 hasConcept C190502265 @default.
- W3166379334 hasConcept C2776214188 @default.
- W3166379334 hasConcept C37736160 @default.
- W3166379334 hasConcept C41008148 @default.
- W3166379334 hasConcept C50644808 @default.
- W3166379334 hasConcept C55493867 @default.
- W3166379334 hasConcept C63479239 @default.
- W3166379334 hasConceptScore W3166379334C104317684 @default.
- W3166379334 hasConceptScore W3166379334C119857082 @default.
- W3166379334 hasConceptScore W3166379334C154945302 @default.
- W3166379334 hasConceptScore W3166379334C185592680 @default.
- W3166379334 hasConceptScore W3166379334C190502265 @default.
- W3166379334 hasConceptScore W3166379334C2776214188 @default.
- W3166379334 hasConceptScore W3166379334C37736160 @default.
- W3166379334 hasConceptScore W3166379334C41008148 @default.
- W3166379334 hasConceptScore W3166379334C50644808 @default.
- W3166379334 hasConceptScore W3166379334C55493867 @default.
- W3166379334 hasConceptScore W3166379334C63479239 @default.
- W3166379334 hasLocation W31663793341 @default.
- W3166379334 hasOpenAccess W3166379334 @default.
- W3166379334 hasPrimaryLocation W31663793341 @default.
- W3166379334 hasRelatedWork W2989784533 @default.
- W3166379334 hasRelatedWork W3091172437 @default.
- W3166379334 hasRelatedWork W3127679336 @default.
- W3166379334 hasRelatedWork W3138091863 @default.
- W3166379334 hasRelatedWork W3175493885 @default.
- W3166379334 hasRelatedWork W4225557503 @default.
- W3166379334 hasRelatedWork W4280534256 @default.
- W3166379334 hasRelatedWork W4287269929 @default.
- W3166379334 hasRelatedWork W4288018014 @default.
- W3166379334 hasRelatedWork W4221145470 @default.
- W3166379334 isParatext "false" @default.
- W3166379334 isRetracted "false" @default.
- W3166379334 magId "3166379334" @default.
- W3166379334 workType "article" @default.