| Abstract: | This work is concerned with smoothed stochastic approximation/optimization algorithms. The main emphasis is placed on the asymptotic optimality issues. An algorithm with averaging in both state variables and observations is studied. Under correlated noise processes, it is shown that a scaled sequence of the iterates converges weakly to a Browman motion. As a result the algorithm is asymptotically optimal Numerical experiments are carried out. Comparisons are made among several algorithms for both linear and nonlinear functions. The numerical results yield good agreement with our analytical findings |
