reinforcement learning advice: Nonlinear Function