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/

Zendy Plus

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

Zendy Tools

Zendy Open

Interpretability is an emerging area of research in trustworthy machinelearning. Safe deployment of machine learning system mandates that theprediction and its explanation be reliable and robust. Recently, it has beenshown that the explanations could be manipulated easily by adding visuallyimperceptible perturbations to the input while keeping the model's predictionintact. In this work, we study the problem of attributional robustness (i.e.models having robust explanations) by showing an upper bound for attributionalvulnerability in terms of spatial correlation between the input image and itsexplanation map. We propose a training methodology that learns robust featuresby minimizing this upper bound using soft-margin triplet loss. Our methodologyof robust attribution training (\textit{ART}) achieves the new state-of-the-artattributional robustness measure by a margin of $\approx$ 6-18 $\%$ on severalstandard datasets, ie. SVHN, CIFAR-10 and GTSRB. We further show the utility ofthe proposed robust training technique (\textit{ART}) in the downstream task ofweakly supervised object localization by achieving the new state-of-the-artperformance on CUB-200 dataset.

Attributional Robustness Training using Input-Gradient Spatial Alignment