Geometry_of_Neural_Reinforcement_Learning_in_Continuous_State_and_Action_Spaces