A heuristic algorithm for stochastic root finding

We propose a heuristic algorithm for n-dimensional stochastic root-finding problems (SRFPs), solving a system of n equations using only the function estimates. This algorithm applies the deterministic Broyden's method into the framework of retrospective approximation, a generic SRFP algorithm. Simulation results for two tandem-queue examples are shown. Our algorithm seems to converge in these examples. The computational time increases substantially with the dimension n.