Robust On-Line Reinforcement Learning Using Kernels. door Hicham Mansouri