Usage of Preudoinversion and Regularization Methods in Adaptive Models of Structural Representation and Data Processing

摘要

Mathematical models of presentation of data in structural form in static case, which are based on direct and dual problems of regularization and estimation of computational complexity of the corresponding methods, are given. Iteration schemes for correction of vector of unknown parameters in the case of dynamical entering of experimental data are developed on their basis. Estimates of convergence of the suggested iterative schemes in the class of ellipsoids, where phase restrictions are set in the form of parallelepiped, are obtained.