Anaphiyev A. Template Theory in the Problems of Learning by Precedents and Model Selection

3. The thesis is devoted to the development of new approach based on the notion of templates which are the operators used to reduce the domain of uncertainty while constructing decision rules in the problems of learning by precedents; the analysis of polynomial template properties used for the problems of learning by precedents and model selection; the study of the properties of complete and partial kernels, basis and fields of regularities, used for optimal algorithm choice; the substantiation of the influence of the validity initial data in the decision-making process. The requirement in the formalization of the uncertainty domain reducing process is grounded for the problems of learning by precedents and model selection.