<ul><li>The paper proposes a nested stochastic gradient descent algorithm for solving regularized nonconvex Distributionally Robust Optimization (DRO) problems.</li><li>The algorithm is designed to handle DRO problems with generalized Sinkhorn distance and nonconvex, unbounded loss functions.</li><li>The proposed algorithm has polynomial iteration and sample complexities that are independent of data size and parameter dimension.</li><li>Numerical experiments demonstrate the efficiency and robustness of the algorithm.</li></ul>

Nested Stochastic Gradient Descent for (Generalized) Sinkhorn Distance-Regularized Distributionally Robust Optimization

Discover more