Abstract:
Iterative stochastic algorithms are considered where the deterministic component is linear. Estimates of the matrix of the process r.m.s. error are given for different ways to select the step length. The results are applied in the study of the gradient method of minimizing the quadratic function in the presence of noises.