Generation and Optimization of Fuzzy Neural Network Structure

Abstract

The paper presents a possibility of exploitation of distributed genetic algorithms (DGAs) for optimization of the neural networks (NNs) and fuzzy neural networks (FNNs) structure and its application to pattern recognition. Generally, there can be several approaches to generation structure of NNs based on genetic algorithms (GAs). Two of them are used most frequently. In the first approach, NNs are only generated from a genotype while in the second approach two genotypes are used. These methods make use of GAs to determine: synapse weights NNs, where their structure is known in advance; NNs structure and synapse weights. This proposal belongs to the second group of methods. These make use of GAs to determine structure and synapse weights NNs (FNNs).