Adaptive Regularization of Representation Rank as an Implicit Constraint of Bellman Equation

Adaptive regularization of the representation rank of value networks in deep reinforcement learning based on the implicit constraints imposed by the Bellman equation.