FAST NUMERICAL NEURAL NETWORK AND FUZZY METHODS FOR STOCHASTIC ESTIMATION OF THE DYNAMIC SYSTEMS STATE
The present work deals with fast numerical methods of stochastic approximation for the problem solution of estimation the dynamic systems state with decomposition. Decomposition structures are realized as growing memory filters on the basis of artificial neural networks, fuzzy systems and their combinations. By the example of solution of nonlinear estimation problem it is shown that by means of synthetic decomposition systems we come up to the high estimation accuracy. It is important that the training speed of these systems is considerably above the training speed of the initial systems without decomposition. The consistency of speedup by using decomposition is in the results.
Keywords: estimation, synthetic system, decomposition, artificial neural network, nonlinear feedforward neural network, radial basis function network, fuzzy system, ANFIS
