A decomposition fuzzy neural network optimization method and device

Abstract

The invention belongs to the technical field of computing intelligent network optimization, and particularly relates to a decomposition fuzzy neural network optimization method and device. The methodcomprises the following steps: establishing a decomposition fuzzy neural network, carrying out iterative clustering on an input data space according to user-specified parameters, obtaining a clustering result, determining the number of components of the decomposition fuzzy neural network according to the clustering result, and decomposing a rule prefix membership function center of the fuzzy neural network; Determining network parameters according to the number of decomposed fuzzy neural network components and a decomposed fuzzy neural network rule precursor membership function center; And determining an optimized decomposition fuzzy neural network according to the network parameters. According to the method, the number of component rules in the decomposition fuzzy neural network is determined through fuzzy clustering; The fuzzy neural network input fuzzy membership function parameter, the rule precursor membership function parameter and the component weight are optimized, so that themodeling precision of the decomposition fuzzy neural network model is higher, the learning time is shortened, the model training and learning efficiency is improved, and the software and hardware expenditure is reduced.

Description

technical field [0001] The invention belongs to the technical field of computing intelligent neural network optimization, in particular to a method and device for decomposing fuzzy neural network optimization. Background technique [0002] In 1965, the concept of fuzzy sets was proposed, marking the birth of a new method capable of describing uncertain phenomena in nature. On this basis, the fuzzy reasoning system was proposed and successfully applied in the fields of complex system modeling and control. However, the traditional fuzzy system is based on the theoretical paradigm of expert system, and its adaptive learning ability is insufficient. In 1995, combining fuzzy theory and neural network, the fuzzy neural network model was proposed, which endowed the neural network with interpretability and fuzzy system learnability. In the learning process of the fuzzy neural network, since the linguistic value of each linguistic variable appears in different rules, in the process...

AI Technical Summary

Problems solved by technology

[0003] For this reason, the present invention provides a method and device for optimizing parameters of a decomposed fuzzy neural network, which solves the problems of the current decomposed fuzzy system, such as a large number of rules, a large amount of calculation, and a large cost. The modeling accuracy is higher, and the network model structure is more concise. Shorten learning time, improve model training and learning efficiency, and reduce software and hardware overhead

Images