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

Classifiers used in the wild, in particular for safety-critical systems,should not only have good generalization properties but also should know whenthey don't know, in particular make low confidence predictions far away fromthe training data. We show that ReLU type neural networks which yield apiecewise linear classifier function fail in this regard as they produce almostalways high confidence predictions far away from the training data. For boundeddomains like images we propose a new robust optimization technique similar toadversarial training which enforces low confidence predictions far away fromthe training data. We show that this technique is surprisingly effective inreducing the confidence of predictions far away from the training data whilemaintaining high confidence predictions and test error on the originalclassification task compared to standard training.

Why ReLU networks yield high-confidence predictions far away from the training data and how to mitigate the problem