This program studies the foundations of machine learning models and their properties, as well as the interactions between data-driven and analytical models in, for example, the efficient simulation of complex physical processes..
Topics include robustness, optimization, verification of performance guarantees, proposing a hybrid AI approach for accelerating the simulation of physical models, and designing and validating certifiable architectures of critical autonomous systems (e.g. in aeronautics and automotive sectors).
In particular, these research subjects are dealt with in part within the framework of the collaborative research project. DEEL (DEpendable Explainable Learning) is a collaborative research project of ANITI that aims to develop technological building blocks toward dependable, robust, explainable and certifiable AI critical systems. IVADO IVADO (The Institute for Data Valorization) and CRIAQ (Consortium for research and innovation in aerospace in Quebec), with the aim of developing technological building blocks allowing the implementation of critical AI systems that are reliable, robust, explainable and certifiable.
The themes associated with the Certifiable AI program :
- Safe design and embeddability
- Fair Learning
- AI & physical models
- Optimisation and game theory for IA
- Automated reasoning and decision making