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

Deep Neural Networks (DNNs) are expected to provide explanation for users tounderstand their black-box predictions. Saliency map is a common form ofexplanation illustrating the heatmap of feature attributions, but it suffersfrom noise in distinguishing important features. In this paper, we propose amodel-agnostic learning method called Saliency Constrained Adaptive AdversarialTraining (SCAAT) to improve the quality of such DNN interpretability. Byconstructing adversarial samples under the guidance of saliency map, SCAATeffectively eliminates most noise and makes saliency maps sparser and morefaithful without any modification to the model architecture. We apply SCAAT tomultiple DNNs and evaluate the quality of the generated saliency maps onvarious natural and pathological image datasets. Evaluations on differentdomains and metrics show that SCAAT significantly improves the interpretabilityof DNNs by providing more faithful saliency maps without sacrificing theirpredictive power.

SCAAT: Improving Neural Network Interpretability via Saliency Constrained Adaptive Adversarial Training