FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2996690718> ?p ?o ?g. }

• W2996690718 abstract "Previous work shows that adversarially robust generalization requires larger sample complexity, and the same dataset, e.g., CIFAR-10, which enables good standard accuracy may not suffice to train robust models. Since collecting new training data could be costly, we focus on better utilizing the given data by inducing the regions with high sample density in the feature space, which could lead to locally sufficient samples for robust learning. We first formally show that the softmax cross-entropy (SCE) loss and its variants convey inappropriate supervisory signals, which encourage the learned feature points to spread over the space sparsely in training. This inspires us to propose the Max-Mahalanobis center (MMC) loss to explicitly induce dense feature regions in order to benefit robustness. Namely, the MMC loss encourages the model to concentrate on learning ordered and compact representations, which gather around the preset optimal centers for different classes. We empirically demonstrate that applying the MMC loss can significantly improve robustness even under strong adaptive attacks, while keeping state-of-the-art accuracy on clean inputs with little extra computation compared to the SCE loss." @default.
• W2996690718 created "2019-12-26" @default.
• W2996690718 creator A5003713785 @default.
• W2996690718 creator A5029335824 @default.
• W2996690718 creator A5031863731 @default.
• W2996690718 creator A5033761498 @default.
• W2996690718 creator A5068755794 @default.
• W2996690718 creator A5080298585 @default.
• W2996690718 date "2020-04-30" @default.
• W2996690718 modified "2023-09-25" @default.
• W2996690718 title "Rethinking Softmax Cross-Entropy Loss for Adversarial Robustness" @default.
• W2996690718 cites W1836465849 @default.
• W2996690718 cites W1932198206 @default.
• W2996690718 cites W1980287119 @default.
• W2996690718 cites W2059081634 @default.
• W2996690718 cites W2108598243 @default.
• W2996690718 cites W2112796928 @default.
• W2996690718 cites W2144172034 @default.
• W2996690718 cites W2167732364 @default.
• W2996690718 cites W2180612164 @default.
• W2996690718 cites W2243397390 @default.
• W2996690718 cites W2302255633 @default.
• W2996690718 cites W2512973121 @default.
• W2996690718 cites W2520774990 @default.
• W2996690718 cites W2679033717 @default.
• W2996690718 cites W2737691244 @default.
• W2996690718 cites W2766557427 @default.
• W2996690718 cites W2773726006 @default.
• W2996690718 cites W2774644650 @default.
• W2996690718 cites W2784293261 @default.
• W2996690718 cites W2791953061 @default.
• W2996690718 cites W2795727551 @default.
• W2996690718 cites W2803392236 @default.
• W2996690718 cites W2906186365 @default.
• W2996690718 cites W2912070915 @default.
• W2996690718 cites W2913848079 @default.
• W2996690718 cites W2955488837 @default.
• W2996690718 cites W2962729158 @default.
• W2996690718 cites W2962872506 @default.
• W2996690718 cites W2962898354 @default.
• W2996690718 cites W2962943487 @default.
• W2996690718 cites W2962972504 @default.
• W2996690718 cites W2963062382 @default.
• W2996690718 cites W2963143631 @default.
• W2996690718 cites W2963166243 @default.
• W2996690718 cites W2963207607 @default.
• W2996690718 cites W2963224870 @default.
• W2996690718 cites W2963466847 @default.
• W2996690718 cites W2963467071 @default.
• W2996690718 cites W2963496101 @default.
• W2996690718 cites W2963540169 @default.
• W2996690718 cites W2963542245 @default.
• W2996690718 cites W2963564844 @default.
• W2996690718 cites W2963565751 @default.
• W2996690718 cites W2963656735 @default.
• W2996690718 cites W2963770044 @default.
• W2996690718 cites W2963777966 @default.
• W2996690718 cites W2963849784 @default.
• W2996690718 cites W2963857521 @default.
• W2996690718 cites W2963894448 @default.
• W2996690718 cites W2964153729 @default.
• W2996690718 cites W2964253222 @default.
• W2996690718 cites W2969985801 @default.
• W2996690718 cites W2970115835 @default.
• W2996690718 cites W2970316625 @default.
• W2996690718 cites W2970317235 @default.
• W2996690718 cites W2970680991 @default.
• W2996690718 cites W2971316968 @default.
• W2996690718 cites W2983219069 @default.
• W2996690718 cites W3039083618 @default.
• W2996690718 cites W3101227480 @default.
• W2996690718 cites W3118608800 @default.
• W2996690718 cites W9657784 @default.
• W2996690718 hasPublicationYear "2020" @default.
• W2996690718 type Work @default.
• W2996690718 sameAs 2996690718 @default.
• W2996690718 citedByCount "28" @default.
• W2996690718 countsByYear W29966907182019 @default.
• W2996690718 countsByYear W29966907182020 @default.
• W2996690718 countsByYear W29966907182021 @default.
• W2996690718 countsByYear W29966907182022 @default.
• W2996690718 crossrefType "proceedings-article" @default.
• W2996690718 hasAuthorship W2996690718A5003713785 @default.
• W2996690718 hasAuthorship W2996690718A5029335824 @default.
• W2996690718 hasAuthorship W2996690718A5031863731 @default.
• W2996690718 hasAuthorship W2996690718A5033761498 @default.
• W2996690718 hasAuthorship W2996690718A5068755794 @default.
• W2996690718 hasAuthorship W2996690718A5080298585 @default.
• W2996690718 hasConcept C104317684 @default.
• W2996690718 hasConcept C106301342 @default.
• W2996690718 hasConcept C11413529 @default.
• W2996690718 hasConcept C119857082 @default.
• W2996690718 hasConcept C121332964 @default.
• W2996690718 hasConcept C153180895 @default.
• W2996690718 hasConcept C154945302 @default.
• W2996690718 hasConcept C167981619 @default.
• W2996690718 hasConcept C185592680 @default.
• W2996690718 hasConcept C188441871 @default.
• W2996690718 hasConcept C1921717 @default.
• W2996690718 hasConcept C2988382989 @default.

• next