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

Constructing accurate model-agnostic explanations for opaque machine learningmodels remains a challenging task. Classification models for high-dimensionaldata, like images, are often inherently complex. To reduce this complexity,individual predictions may be explained locally, either in terms of a simplerlocal surrogate model or by communicating how the predictions contrast withthose of another class. However, existing approaches still fall short in thefollowing ways: a) they measure locality using a (Euclidean) metric that is notmeaningful for non-linear high-dimensional data; or b) they do not attempt toexplain the decision boundary, which is the most relevant characteristic ofclassifiers that are optimized for classification accuracy; or c) they do notgive the user any freedom in specifying attributes that are meaningful to them.We address these issues in a new procedure for local decision boundaryapproximation (DBA). To construct a meaningful metric, we train a variationalautoencoder to learn a Euclidean latent space of encoded data representations.We impose interpretability by exploiting attribute annotations to map thelatent space to attributes that are meaningful to the user. A difficulty inevaluating explainability approaches is the lack of a ground truth. We addressthis by introducing a new benchmark data set with artificially generated Irisimages, and showing that we can recover the latent attributes that locallydetermine the class. We further evaluate our approach on tabular data and onthe CelebA image data set.

Explaining Predictions by Approximating the Local Decision Boundary