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

Explainability for machine learning models has gained considerable attentionwithin the research community given the importance of deploying more reliablemachine-learning systems. In computer vision applications, generativecounterfactual methods indicate how to perturb a model's input to change itsprediction, providing details about the model's decision-making. Currentmethods tend to generate trivial counterfactuals about a model's decisions, asthey often suggest to exaggerate or remove the presence of the attribute beingclassified. For the machine learning practitioner, these types ofcounterfactuals offer little value, since they provide no new information aboutundesired model or data biases. In this work, we identify the problem oftrivial counterfactual generation and we propose DiVE to alleviate it. DiVElearns a perturbation in a disentangled latent space that is constrained usinga diversity-enforcing loss to uncover multiple valuable explanations about themodel's prediction. Further, we introduce a mechanism to prevent the model fromproducing trivial explanations. Experiments on CelebA and Synbols demonstratethat our model improves the success rate of producing high-quality valuableexplanations when compared to previous state-of-the-art methods. Code isavailable at https://github.com/ElementAI/beyond-trivial-explanations.

Beyond Trivial Counterfactual Explanations with Diverse Valuable Explanations