This is the authors' abstract. We don't add key points for this paper.
In order to evaluate whether a test set for deep neural networks (DNNs) is adequate, people proposed many test adequacy metrics for DNNs testing. As an enhancement of the classic fixed strength combinatorial testing technique, variable strength combinatorial testing has also been used for white-box testing of DNNs to evaluate whether the activation behavior of neurons is diverse during test execution. However, it is still inconclusive whether the variable strength combi-natorial coverage of neurons is beneficial to the testing of DNNs. We conduct an empirical study in this paper to answer how the variable strength combinatorial coverage of neurons affects the fault detection in DNNs. The experimental results show a medium or strong negative correlation between the variable strength combinatorial coverage of neurons and the number of adversarial examples for DNNs. This may imply that adversarial examples generated using the same adversarial attack method have similarity in the characteristics of neural output.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Huang et al. (2024) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: