Algorithms for Mathematical Remodeling and Generation of Training Sets

In this paper the problem of mathematical remodeling is considered. The conception of a new model building based on the already known model belonging to another class is described. It is shown that the central task is the generation of a training set, which can be produced in different ways. Algorithm for generation of data based on accounting for ranges of variation of the original function using interval estimation methods is proposed. The proposed approach consists of sequence of box bisections and the random points generation in proportion to the ranges of function range changes. The generalized mathematical remodeling algorithm is presented in the paper.