Research On Fuzzy Neural Network Models And Their Application

Artificial intelligence is an advanced technology and has become the focus of international competition.Fuzzy theory and fuzzy neural network as an important research content of uncertain artificial intelligence have been applied successfully in the fields of intelligent control,pattern recognition,data analysis and intelligent information processing.Among them,The method of determining membership function has not been solved very well,and the design of effective learning algorithm of fuzzy neural network has been the difficulty and hot spot of research.The main theory,research and system development works in this thesis are as follows:1.In engineering practice,it is often necessary to transform a fuzzy set expressed on a finite set domain into a fuzzy set expressed on an interval domain.Therefore,it is proposed an algorithm to determine the membership functions of multiple fuzzy sets,which can be unified and coordinated and ensure the minimum error,and gives the corresponding mathematical principle in this thesis.This algorithm improves the existing method that only one finite set can be transformed into a fuzzy set at a time.2.Based on Dombi's t-norm TD,a fuzzy associative memory network model with parameters ?,Max_TD FAM,is proptosed,and its analytical and effective learning algorithm is designed.The learning algorithm is efficient because it does not require iteration,not require small increment,not require loop,and can directly give the specific value of the connection weights.Using this learning algorithm to train Max_TD FAM can make the network have complete and reliable recall ability.By selecting different parameters ?,the corresponding Max_TD FAM network can conform to more requirements in practice.3.A water basin data management and intelligent analysis system is designed and implemented in accordance with software engineering specifications.The system uses the Max_TD FAM model to predict some water data and presents a variety of data analysis results in a graphical manner.The system can run stably and has certain practicability.