Global optimality of softmax policy gradient with single hidden layer neural networks in the mean-field regime

Open in new window