Stochastic approximation methods in nonlinear regression models 1

The brief review of theory of stochastic approximation methods in nonlinear regression models is given. Particularly the situation when regression function has several extrema is discussed. The special attention is given to recent on the weakening of conditions on distrubances and on regression function, which guarantee the convergrence of the provedures. The problems of asymptotically optimal choice of procedures, parameters and applications to the parametric estimation are touched too.