An extension of the Kiefer-Wolfowitz stochastic approximation procedure
Y. Maeda · 2003
The Kiefer-Wolfowitz stochastic approximation procedure is utilized to find a maximum or a minimum point of a regression function. In the present work, an algorithm that is an extension of the usual Kiefer-Wolfowitz stochastic approximation procedure is proposed. The algorithm corresponds to an adaptive version of the usual Kiefer-Wolfowitz stochastic approximation procedure. The convergence of this algorithm is proved. The proposed algorithm has a faster convergence rate than the usual Kiefer-Wolfowitz procedure. A numerical simulation is shown.>