Evolving neuro-fuzzy systems with kernel activation functions

The aim of this book is to develop new methods for adaptive learning for evolving neural networks and neuro-fuzzy systems with kernel activation functions. The book provides an overview on different principles of neural networks' learning, basic and most popular neural networks and neuro-fuzzy systems where kernel constructions are used as activation functions. Advantages and shortcomings are defined for well-known approaches. The proposed evolving architectures can be used for time series' and data streams' processing. Taking into consideration all the mentioned above, we proposed an adaptive method of ensembles' tuning for neural networks with kernel activation functions which are learnt on both optimization procedures and memory (that gives an optimal accuracy of an ensemble output signal). Speaking of processing data streams in an online mode, we introduced a method for ensembles' tuning for neuro-fuzzy systems with kernel activation functions which are learnt on both optimization procedures and memory (that provides a high accuracy of an ensemble output signal on the basis of fuzzy generalization).

Autorentext

Yevgeniy V. Bodyanskiy is the Professor of Artificial Intelligence Department at Kharkiv National University of Radio Electronics (Ukraine) and the Scientific Head in Control Systems Research Laboratory. He has more than 500 scientific publications (40 inventions and 10 monographs). He's interested in hybrid systems of Computational Intelligence.