A stochastic successive minimization method for nonsmooth nonconvex optimization with applications to transceiver design in wireless communication networks (Q301668): Difference between revisions

The authors consider the following optimization problem \[ \min_{.}f\left( x\right) \triangleq \mathbb{E}_{\xi }\left[ g\left( x,\xi \right) \right]\text{s.t.}\;\qquad x\in \chi, \eqno(1) \] where \(\chi \subseteq \mathbb{R}^{n}\) is a bounded closed convex set; \(\xi \) is a random vector drawn from a set \(\Xi \in\mathbb{R}^{m}\), \(g:\chi \times \Xi \longmapsto \mathbb{R}\) is a real-valued function and \(\mathbb{E}\left( \cdot \right) \) represents the expectation operator. The classical sample average approximation method for solving this problem requires the minimization of an ensemble average of the objective at each step, which can be expensive. In this paper, the authors propose a stochastic successive upper-bound minimization method (SSUM) which minimizes an approximate ensemble average at each iteration. The main contributions of this paper include the development and analysis of the SSUM method as well as its applications in linear transceiver design for wireless communication networks and online dictionary learning. Using the SSUM framework, the authors extend the classical stochastic (sub-)gradient method to the problem of minimizing a nonsmooth nonconvex objective function and establish its covergence.