Softmax Deep Double Deterministic Policy Gradients