Recent work has shown that deep neural networks are capable ofapproximating both value functions and policies in reinforcementlearning domains featuring continuous state and ... International ...