Format des notes

Programme détaillé

The first part of the course introduces the fundamental concepts of formal verification, providing students with the theoretical foundations needed to understand how the correctness of systems can be formally established. It then focuses on model checking for closed systems, exploring techniques for verifying system properties based on formal models. The course subsequently addresses model checking for multi-agent systems, considering the additional challenges that arise when multiple agents interact and make decisions within a shared environment. Finally, it examines the computational complexities of model checking, discussing the resources required to verify different types of systems and properties, as well as the main challenges associated with the scalability of verification techniques.

The second part of the course focuses on multi-agent reinforcement learning (MARL) from a practical perspective, examining how agents learn to make decisions through interaction with a shared environment. It introduces the core concepts and main algorithmic approaches used in cooperative, competitive, and mixed settings, with an emphasis on their implementation and application. Students gain hands-on experience designing environments, specifying observations and rewards, and training and evaluating multi-agent learning algorithms. Through experiments, students explore how modelling choices and training procedures influence performance, learning to assess the strengths and limitations of MARL methods in practice.