Nonparametric modeling of multidimensional memoryless processes

Abstract

The report considers the case when multidimensional memoryless objects have an unknown stochastic dependence between output variables, a training sample is available. Such processes are called T-processes. Constructing a model for such a process leads to solve a system of implicit functions. Moreover, the form of these functions is unknown up to parameters. Therefore, practical application of generally accepted parametric identification theory is not possible. In this case, we will use a piecemeal approach to predict output variables from known input variables