Methodological Guideline and Evaluation tools for robust artificial intelligence

In this document, we present the work that has been done during the second year of the Confiance.ai program on the methods and tools related to robustness of artificial intelligence models and their applications to real industrial problems proposed by the program industrial partners.
The following document details the methodological progress and numerical results obtained in 6 research and development topics that are strongly interconnected. As an introduction we propose to the reader a short overview of each section.