Complexity of families of learning algorithms and estimation of the nonrandomness of extraction of empirical regularities

The paper presents a general approach to the evaluation of the complexity of classes of algorithms, so-called pVCD-method. To develop this method, all the examined families of models of empiric generalization were restricted to classes implementable on computers and, wider, by examining their partially recursive representations. Within the framework of the algorithmic approach, the concept of Kolmogorov’ complexity of algorithms for the recognition of properties or the extraction of regularities is proposed. The method proposed to evaluate the nonrandomness of the extraction of empirical regularities is based on this concept.