Model-Based Reinforcement Learning by Pyramidal Neurons : Robustness of the Learning Rule