An Approach to Granular Testing of Deep Neural Networks
Establishing rigorous testing and certification techniques is essential before deploying new technologies like Deep Learning (DL) in safety-critical applications. We propose a testing approach that could identify weaknesses in DL models.