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

Batch normalization (BN) has been widely used in modern deep neural networks(DNNs) due to improved convergence. BN is observed to increase the modelaccuracy while at the cost of adversarial robustness. There is an increasinginterest in the ML community to understand the impact of BN on DNNs, especiallyrelated to the model robustness. This work attempts to understand the impact ofBN on DNNs from a non-robust feature perspective. Straightforwardly, theimproved accuracy can be attributed to the better utilization of usefulfeatures. It remains unclear whether BN mainly favors learning robust features(RFs) or non-robust features (NRFs). Our work presents empirical evidence thatsupports that BN shifts a model towards being more dependent on NRFs. Tofacilitate the analysis of such a feature robustness shift, we propose aframework for disentangling robust usefulness into robustness and usefulness.Extensive analysis under the proposed framework yields valuable insight on theDNN behavior regarding robustness, e.g. DNNs first mainly learn RFs and thenNRFs. The insight that RFs transfer better than NRFs, further inspires simpletechniques to strengthen transfer-based black-box attacks.

Batch Normalization Increases Adversarial Vulnerability and Decreases Adversarial Transferability: A Non-Robust Feature Perspective