Volume no :16, Issue no: 2, October (2017)

Journal of Materials Science and Engineering with Advanced Technology

HIGH-PERFORMANCE METHOD FOR CONSTRUCTING MEMBERSHIP FUNCTIONS BASED ON THE FUZZY CELLULAR SELF-ORGANIZING NEURAL NETWORKS

Author's: S. V. Gorbachev
Pages: [53] - [68]
Received Date: March 26, 2017
Submitted by:
DOI: http://dx.doi.org/10.18642/jmseat_7100121808

Abstract

The article is devoted to the actual questions of the initialization of membership functions of fuzzy sets in adaptive systems analysis and fuzzy inference. Presents an analysis of methods for constructing such functions. Proposed high-performance neural network method based on new hybrid structures - fuzzy self-organizing Kohonen cellular neural networks (FCNN-SOM), allowing to automatize the process of fuzzification of the input variables and characterized by a high degree of self-organization of neurons, the lack of heuristic parameters of learning, the ability to selectively control individual connections between neurons to solve the problem of “dead” neurons and improved separating properties of the network in the case of overlapping clusters.

Keywords

technology, innovative projects, expert data, membership functions, fuzzy self-organizing cellular networks.