A Regularized Opponent Model with Maximum Entropy Objective