The challenges of model-based reinforcement learning and how to overcome them - Csaba Szepesvári