Stochastic Approximation
Alan J. Gross, David C. McLean · Encyclopedia of Biostatistics · 2005
Abstract Stochastic approximation is a method for approximating the root of the equationM(θ) = m0, when the functionM(.) is not directly observable but can be cast as the expectation of a suitably defined regression function. The method was introduced by Robbins and Munro in 1951 and has since been extensively refined and extended. It has found application in many areas of biostatistics and in the application of neural networks.