I\'m trying to use Keras to run a reinforcement learning algorithm. In this algorithm, I\'m training a neural network. What\'s different from other learning problems is that