Popov S. Adaptive methods of stochastic observation fields processing

Problems of adaptive filtering, smoothing, prediction of a wide class of stochastic observation fields under uncertainty conditions in real time are solved with the aim of processing effectiveness increase using adaptive systems and artificial neural networks theories. New adaptive method for 2-D models' parameters estimation is proposed and its convergence is proved. New adaptive and neural network methods for polyharmonic fields decomposition are obtained that does not require a priori information about their structure and parameters. The problem of adaptive processing of observation fields sequences is solved on the basis of a simplified matrix model using new single-step optimal adaptive method and multistep method of RLS type for its parameters tuning. New architectures of artificial neural networks for adaptive processing of stochastic observation fields sequences are developed. The proposed methods have been used for solving real-world problems of heat fields prediction and economic data smoothi ng