Approximation of continuous functions with random errors in observed values
Let a continuous function \(f(t)\), \(t\in [0,1]\), be defined only at a finite number of points \(\tau_{i}\leq 1\) as a sum of true values and random mutually independent normally distributed values \(\xi_{i}: f(\tau_{i}) + \xi_{i}\). The author proposes a procedure of \(m\)-th approximation of the function \(f(t)\) by a method which is similar to the Galerkin method. This approximation and the error of the approximation also have normal distributions. The purpose is to calculate confidence limits for the error of approximation simultaneously for all functions \(f\) from a given class of functions \(X_{0}\). The class of functions \(X_{0}\) is defined by the condition of uniform boundedness of the error of approximation. In the article existence of such confidence limits of the given significance level is proved and a procedure of its calculation is determined. An example of applications of the result when the approximation is done with the help spline functions is considered.
- On approximation of functions by random polynomials
- On the approximation of function by random splines of minimal defect
- Approximation of random processes by cubic splines
- Weighted approximation of random functions
- Monotone approximation of random functions with multivariate domains in respect of lattice semi-norms
This page was built for publication: Approximation of continuous functions with random errors in observed values
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4705221)
