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

Modern software systems rely on Deep Neural Networks (DNN) when processingcomplex, unstructured inputs, such as images, videos, natural language texts oraudio signals. Provided the intractably large size of such input spaces, theintrinsic limitations of learning algorithms, and the ambiguity about theexpected predictions for some of the inputs, not only there is no guaranteethat DNN's predictions are always correct, but rather developers must safelyassume a low, though not negligible, error probability. A fail-safe DeepLearning based System (DLS) is one equipped to handle DNN faults by means of asupervisor, capable of recognizing predictions that should not be trusted andthat should activate a healing procedure bringing the DLS to a safe state. Inthis paper, we propose an approach to use DNN uncertainty estimators toimplement such a supervisor. We first discuss the advantages and disadvantagesof existing approaches to measure uncertainty for DNNs and propose novelmetrics for the empirical assessment of the supervisor that rely on suchapproaches. We then describe our publicly available tool UNCERTAINTY-WIZARD,which allows transparent estimation of uncertainty for regular tf.keras DNNs.Lastly, we discuss a large-scale study conducted on four different subjects toempirically validate the approach, reporting the lessons-learned as guidancefor software engineers who intend to monitor uncertainty for fail-safeexecution of DLS.

Fail-Safe Execution of Deep Learning based Systems through Uncertainty Monitoring