scientific article; zbMATH DE number 177230

From MaRDI portal
Publication:4037080





[For the entire collection see Zbl 0754.00005.] Let \(f(x)\) be an unknown function of \(x\in\mathbb R^ N\) having a unique minimizer \(x^*\). The problem under consideration is to estimate \(x^*\) from the observations \(y_ i=f(x_ i)+\xi_ i\), \(i=1,2,\dots,n\), where \(x_ i\)'s are the design points and \(\xi_ i\) are certain errors. The authors studied randomized procedures for estimating \(x^*\) in their earlier work [Probl. Inf. Transm. 26, No. 2, 126--133 (1990); translation from Probl. Peredachi Inf. 26, No. 2, 45--53 (1990; Zbl 0712.90073)]. Here, the authors establish the convergence of their randomized procedures under very general and weak assumptions on \(\xi_ i\), namely, \(\xi_ i\)'s are allowed to be nonrandom or random with arbitrary dependence. They show that the randomized procedures converge at an optimal rate under such conditions on \(\xi_ i\).











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4037080)