Learning RoboCup-Keepaway with Kernels

We give another success story of using kernel-based methods to solve a dificult reinforcement learning problem, namely that of 3vs2 keepaway in RoboCup simulated soccer. Key challenges in keepaway are