CONCERNING SOME PROCEDURES OF FUNCTION APPROXIMATION BASED ON THE USE OF EXPERIMENTAL DATA.

Some problems connected with function approximation, based on experimental data, are considered. Some of the problems investigated concern the difference between ″good″ and ″bad″ expansion, a criterion for selecting the basis functions for expansion, and some algorithmical approaches to improve the quality of approximation. In the last part of the paper a comparative analysis of some procedures of function approximation based on the use of experimental data is performed. The advantage of the least mean square approximation in comparison with some other procedures is ascertained, although the latter procedures would have been supposed to yield better results since they make use of some additional information. An example demonstrating the result is considered.