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

When Deep Neural Networks (DNNs) are used in safety-critical systems,engineers should determine the safety risks associated with failures (i.e.,erroneous outputs) observed during testing. For DNNs processing images,engineers visually inspect all failure-inducing images to determine commoncharacteristics among them. Such characteristics correspond tohazard-triggering events (e.g., low illumination) that are essential inputs forsafety analysis. Though informative, such activity is expensive anderror-prone.  To support such safety analysis practices, we propose SEDE, a technique thatgenerates readable descriptions for commonalities in failure-inducing,real-world images and improves the DNN through effective retraining. SEDEleverages the availability of simulators, which are commonly used forcyber-physical systems. It relies on genetic algorithms to drive simulatorstowards the generation of images that are similar to failure-inducing,real-world images in the test set; it then employs rule learning algorithms toderive expressions that capture commonalities in terms of simulator parametervalues. The derived expressions are then used to generate additional images toretrain and improve the DNN.  With DNNs performing in-car sensing tasks, SEDE successfully characterizedhazard-triggering events leading to a DNN accuracy drop. Also, SEDE enabledretraining leading to significant improvements in DNN accuracy, up to 18percentage points.

Simulator-based explanation and debugging of hazard-triggering events in DNN-based safety-critical systems