Sequential Estimation in Stochastic Approximation Problem with Autoregressive Errors in Observations

From MaRDI portal





Stochastic approximation methods are known as iterative algorithms for finding a root of a function in the case when functional values can be determined only with a random error. These methods are employed in many applications. Mathematical properties of the above mentioned methods have been investigated, first, under the assumption that the observed errors can be modelled as a sequence of independent identically distributed random variables. Later, the methods have been investigated in the case of stationary processes as a noise, which can describe the situation of correlated observations. The asymptotic theory for the stationary noise is now already well developed. However, other problems have appeared in applications. Especially, there has arisen a serious problem in the case of autoregressive sequences with unknown parameters. It is a problem of uniform convergence of algorithms with respect to these unknown parameters. The aim of the authors is to deal just with this problem. In detail, they construct adaptive iterative procedures which converge uniformly (in some sense) with respect to the nuisance parameters.











This page was built for publication: Sequential Estimation in Stochastic Approximation Problem with Autoregressive Errors in Observations

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4429468)