One-sided convergence of the Robbins-Monro process (Q1098528)

A modified Robbins-Monro process is considered for estimation of roots of regression functions. It is shown that the iterative process converges to the root with probability one and that the estimates exceed the root only by a finite number of times with probability one. The result has been established by the author previously [ibid. 41, 1682-1684 (1981); translation from Avtom. Telemekh. 1980, No.12, 72-75 (1980; Zbl 0464.62081)] but under more stringent assumptions.