High-level Coverage Metrics for Deep Neural Networks

My current work in this field addresses the lack of high-level, semantics-based test coverage metrics for Deep Neural Networks (DNNs). This work is carried out with Xiaowei Huang, in the context of a project funded by the Defence Science and Technology Laboratory (DSTL). In practice, our work mostly revolves around the DeepConcolic toolbox that is dedicated to the analysis of deep neural networks.

Preliminary Paper