Search

We consider the incremental subgradient algorithm employing dynamic step sizes for minimizing the sum of a large number of component convex functions. The dynamic step size rule was firstly introduced by Goffin and Kiwiel [Math. Program., 1999, 85(1): 207‐211] for the subgradient algorithm, soon later, for the incremental subgradient algorithm by Nedic and Bertsekas in [SIAM J. Optim., 2001, 12(1): 109‐138]. It was observed experimentally that the incremental approach has been very successful in solving large separable optimizations and that the dynamic step sizes generally have better computational performance than others in the literature. In the present paper, we propose two modified dynamic step size rules for the incremental subgradient algorithm and analyse the convergence and complexity properties of them. At last, the assignment problem is considered and the incremental subgradient algorithms employing different kinds of dynamic step sizes are applied to solve the problem. The computational experiments show that the two modified ones converges dramatically faster and more stable than the corresponding one in [SIAM J. Optim., 2001, 12(1): 109‐138]. Particularly, for solving large separable convex optimizations, we strongly recommend the second one (see Algorithm 3.3 in the paper) since it has interesting computational performance and is the simplest one. [ABSTRACT FROM AUTHOR]

In this paper, we study smooth stochastic multilevel composition optimization problems, where the objective function is a nested composition of T functions. We assume access to noisy evaluations of the functions and their gradients, through a stochastic first-order oracle. For solving this class of problems, we propose two algorithms using moving-average stochastic estimates, and analyze their convergence to an e-stationary point of the problem. We show that the first algorithm, which is a generalization of [S. Ghadimi, A. Ruszczynski, and M. Wang, SIAM J. Optim., 30 (2020), pp. 960-979] to the T level case, can achieve a sample complexity of OT(1/ε6) by using minibatches of samples in each iteration, where OT hides constants that depend on T. By modifying this algorithm using linearized stochastic estimates of the function values, we improve the sample complexity to OT(1/ε4). This modification not only removes the requirement of having a minibatch of samples in each iteration, but also makes the algorithm parameter-free and easy to implement. To the best of our knowledge, this is the first time that such an online algorithm designed for the (un)constrained multilevel setting obtains the same sample complexity of the smooth single-level setting, under standard assumptions (unbiasedness and boundedness of the second moments) on the stochastic first-order oracle. [ABSTRACT FROM AUTHOR]