Constrained Optimization over Discrete Sets via SPSA with Application to Resource Allocation

摘要

A version of the Simultaneous Perturbation Stochastic Approximation (SPSA) algorithm for optimizing functions over discrete sets is considered. The primary motivation for discrete SPSA is to solve a class of resource allocation problems. The basic algorithm will be described and simulation results presented to illustrate its performance. The application of the algorithm to optimal resource allocation is discussed, wherein the goal is to distribute a finite number of discrete resources to finitely many users in such as way as to optimize a specified objective function.