English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Model compression and model defense for deep neural networks (DNNs) have beenextensively and individually studied. Considering the co-importance of modelcompactness and robustness in practical applications, several prior works haveexplored to improve the adversarial robustness of the sparse neural networks.However, the structured sparse models obtained by the exiting works suffersevere performance degradation for both benign and robust accuracy, therebycausing a challenging dilemma between robustness and structuredness of thecompact DNNs. To address this problem, in this paper, we propose CSTAR, anefficient solution that can simultaneously impose the low-rankness-basedCompactness, high STructuredness and high Adversarial Robustness on the targetDNN models. By formulating the low-rankness and robustness requirement withinthe same framework and globally determining the ranks, the compressed DNNs cansimultaneously achieve high compression performance and strong adversarialrobustness. Evaluations for various DNN models on different datasetsdemonstrate the effectiveness of CSTAR. Compared with the state-of-the-artrobust structured pruning methods, CSTAR shows consistently better performance.For instance, when compressing ResNet-18 on CIFAR-10, CSTAR can achieve up to20.07% and 11.91% improvement for benign accuracy and robust accuracy,respectively. For compressing ResNet-18 with 16x compression ratio on Imagenet,CSTAR can obtain 8.58% benign accuracy gain and 4.27% robust accuracy gaincompared to the existing robust structured pruning method.

CSTAR: Towards Compact and STructured Deep Neural Networks with Adversarial Robustness