Kengy Barty (EDF R&D)
Approximate Dynamic technique often involves the knowledge of a priori concerning the shape of the optimal solution. The field of reinforcement learning is concern by the way to learn about the optimal strategy, using the response of the system under an feedback. The aim of this talk is to recall some results about learning algorithms in an non-parametric scope, afterward we are going to present a promising track to explain the convergence of some non-parametric learning algorithms in Hilbert space.