Estimating functional dependences based on time series with the use of classes of regression functions of infinite capacity

Abstract

The problem of estimating the functional dependence of time series on the time index is considered in the case of short data samples. It is proved that empirical risk functional uniformly converges to the theoretical one in the case where regression functions can be approximated by finite-degree polynomials. An example of estimating the functional dependence for a class of trigonometric functions is presented.