Running the reinforcement learning process