Discrete optimization via SPSA

We consider the use of a fixed gain version of SPSA (simultaneous perturbation stochastic approximation) for optimizing a class of discrete functions. The procedure has been modified to obtain an optimization method that is applicable to cost functions defined on a grid of points in Euclidean p-space having integer components. We discuss some related results on fixed gain SPSA and describe an application of the method to a resource allocation problem.