[EN] Optimal control and reinforcement learning have an associate “value function” which must be suitably approximated. Value function approximation problems usually have different precision requirements in different regions ...