Teslenko N. Self-learning neuro-fuzzy models and systems in the intelligent data analysis tasks

The thesis is devoted to a research of neuro-fuzzy models and systems and their learning and self-learning methods for such intelligent data analysis tasks as functional relations restoring, numeric data attribute space dimensionality reduction and autoassociative memories problems in the sequential processing mode. The model of neural autoaasociative memory on the basis of fuzzy basis functions, that allowed to increase the number of stored patterns and to associate the process of restoring in neural network model with fuzzy clustering procedures; the model of general regression neuro-fuzzy network, that allowed to provide for nonlinear non-stationary objects prediction and identification accuracy and rate increasing; the model of adaptive F-transform, that allowed to process data sequentially and to change the number of membership functions during the process of learning are proposed for the first time. The self-learning methods of autoassociative neural networks for data attribute space dimensionality reduction, that allowed to increase data processing accuracy and rate are developed. Experimental results confirm the efficiency of the proposed methods and models on a number of well-known benchmarks and the real data processing problems.