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

Attribution methods have been shown as promising approaches for identifyingkey features that led to learned model predictions. While most existingattribution methods rely on a baseline input for performing featureperturbations, limited research has been conducted to address the baselineselection issues. Poor choices of baselines limit the ability of one-vs-one(1-vs-1) explanations for multi-class classifiers, which means the attributionmethods were not able to explain why an input belongs to its original class butnot the other specified target class. 1-vs-1 explanation is crucial whencertain classes are more similar than others, e.g. two bird types amongmultiple animals, by focusing on key differentiating features rather thanshared features across classes. In this paper, we present GAN-based ModelEXplainability (GANMEX), a novel approach applying Generative AdversarialNetworks (GAN) by incorporating the to-be-explained classifier as part of theadversarial networks. Our approach effectively selects the counterfactualbaseline as the closest realistic sample belong to the target class, whichallows attribution methods to provide true 1-vs-1 explanations. We showed thatGANMEX baselines improved the saliency maps and led to stronger performance onperturbation-based evaluation metrics over the existing baselines. Existingattribution results are known for being insensitive to model randomization, andwe demonstrated that GANMEX baselines led to better outcome under the cascadingrandomization of the model.

GANMEX: One-vs-One Attributions Guided by GAN-based Counterfactual Explanation Baselines