Genetically Optimized Fuzzy Set-Based Polynomial Neural Networks Based on Information Granules with Aids of Symbolic Genetic Algorithms

Abstract

In this paper, we propose a new architecture of Fuzzy Set–based Polynomial Neural Networks (FSPNN) with a new fuzzy set-based polynomial neuron (FSPN) whose fuzzy rules include the information granules (about the real system) obtained through Information Granulation. Although the conventional FPNN with Fuzzy Relation-based Polynomial Neurons has good approximation ability and generalization capability, there is an important drawback that FPNN is very complicated. If we adopt fuzzy set-based fuzzy rules as substitute for fuzzy relation-based fuzzy rules, we can get an advantage of the rule reduction. We use FSPN as a node of Fuzzy Polynomial Neural Networks to reduce the complexity of the FPNN. The proposed FPNN with Fuzzy Set-based Polynomial Neuron can achieve compactness. Information Granulation can extract good information from numerical data without expert’s knowledge which is important for building Fuzzy Inference System. We put Information Granulation to the proposed FSPN. The structure of the proposed FPNN with FSPN is determined with aids of symbolic gene type genetic algorithms.