Efficient Stochastic Gradient Descent for Learning with Distributionally Robust Optimization