ON STOCHASTIC APPROXIMATION
Hans W. Wolff · ETS Research Bulletin Series · 1970
ABSTRACT This paper deals with a stochastic process for the approximation of the root of a regression equation. This process was first suggested by Robbins and Monro [1]. The main result here is a necessary and sufficient condition on the iteration coefficients for convergence of the process (convergence with probability one and convergence in the quadratic mean).