"Reinforced learning, the basics and the case of asynchronous events" Frédérick Garcia

In the first part of this mini-course, he will present the history and basic concepts of reinforcement learning, a family of methods whose objective is to learn, through trial and error, optimal decision-making strategies in an unfamiliar environment. I will introduce Markov decision processes, time difference methods and Q-learning, the exploration-exploitation dilemma, the notion of traceability and generalization techniques. The second part of the mini-course will be more specifically devoted to the special case of continuous time and asynchronous events, with a focus on STDP in spike neuron models.

Frédérick Garcia is INRA Research Director and Deputy Director of the Convergence Institute Digital Agriculture #DigitAg; MIAT Applied Mathematics and Computer Science Unit of Toulouse.