The neural network is one of the main methods of supervised automatic learning. Verifying its behaviour, i.e. the evolution of its outputs according to the variation of its inputs, is a real challenge.
However, this verification is necessary in an uncertain environment where neural network inputs are noisy. In this article, Mohamed Ibn Kheder (Research Engineer at IRT SystemX) introduces the importance of neural network verification, shares some examples of applications and describes the formulation of their verification.
