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

Adversarial attacks have been proven to be potential threats to Deep NeuralNetworks (DNNs), and many methods are proposed to defend against adversarialattacks. However, while enhancing the robustness, the clean accuracy willdecline to a certain extent, implying a trade-off existed between the accuracyand robustness. In this paper, we firstly empirically find an obviousdistinction between standard and robust models in the filters' weightdistribution of the same architecture, and then theoretically explain thisphenomenon in terms of the gradient regularization, which shows this differenceis an intrinsic property for DNNs, and thus a static network architecture isdifficult to improve the accuracy and robustness at the same time. Secondly,based on this observation, we propose a sample-wise dynamic networkarchitecture named Adversarial Weight-Varied Network (AW-Net), which focuses ondealing with clean and adversarial examples with a ``divide and rule" weightstrategy. The AW-Net dynamically adjusts network's weights based on regulationsignals generated by an adversarial detector, which is directly influenced bythe input sample. Benefiting from the dynamic network architecture, clean andadversarial examples can be processed with different network weights, whichprovides the potentiality to enhance the accuracy and robustnesssimultaneously. A series of experiments demonstrate that our AW-Net isarchitecture-friendly to handle both clean and adversarial examples and canachieve better trade-off performance than state-of-the-art robust models.

Revisiting the Trade-off between Accuracy and Robustness via Weight Distribution of Filters