作者: J.C. Spall
DOI: 10.1109/9.119632
关键词: Minimax approximation algorithm 、 Asymptotic distribution 、 Approximation algorithm 、 Standard algorithms 、 Simultaneous perturbation stochastic approximation 、 Stochastic approximation 、 Mathematical analysis 、 Approximation error 、 Mathematics 、 Function approximation
摘要: The problem of finding a root the multivariate gradient equation that arises in function minimization is considered. When only noisy measurements are available, stochastic approximation (SA) algorithm for general Kiefer-Wolfowitz type appropriate estimating root. paper presents an SA based on simultaneous perturbation instead standard finite-difference Keifer-Wolfowitz procedures. Theory and numerical experience indicate can be significantly more efficient than algorithms large-dimensional problems. >