Bezsonov O. Neural network CMAC and its usage for adaptive processing of information

Analyze of adaptive information processing with using of ANN CMAC was made in the thesis. A biological model of the cerebellum is investigated. A mathematical analogue of this model is a basis for ANN CMAC. The principles of CMAC building and functioning were examined. The structure of CMAC NN was investigated. An analyze of the information coding methods in CMAC was performed. The properties of the different schemes of information coding were investigated. A task of the basis functions choosing in CMAC network was researched. An method of information coding and the basis functions choosing in ANN CMAC which allows to use the network in the tasks of nonlinear neurocontrol without a reference model is proposed. The methods of a training process reduction were developed. The multi-step projection training algorithms were developed and their recurrent forms were received. Imitation modeling of the different training algorithms was performed in MatLab 6.1 environment. A process of solving tasks of identification and control of nonlinear dynamical objects, image and signal filtration and coding was investigated.