Isolation Hierarchical Optimization Algorithm Of MIMO Polygonal Fuzzy Neural Network

A fuzzy neural network is a new type skills combination of an artifical network and fuzzy logic reasoning,it can effectively deal with natural language information and has the strong approximation ability of nonlinear functions.A polygonal fuzzy neural network is a new type of special feedforward network that is based on the special operations of polygonal fuzzy num-bers,by determining limited points of polygonal fuzzy numbers to complete fuzzy information processing,in addition,which has logical reasoning,numerical calculation and has the strong approximation ability of nonlinear functions.At present,the research on polygonal fuzzy neural network is only limited to the single input single output(SISO).Its approximation performance and parameter optimization are still stay in primary state,which brings inconvenience in the network to widely used in further.Therefore,in this paper,the weighting valued parameters of every layer are optimized in multiple input single output(MISO)model and the multi-input and multi-output(MIMO)model.The main research content is divided into following three parts:In the first chapter:it introduces the background,the present research situation,and the preliminaries.In the second chapter:the multi-input and multi-output(MIMO)polygonal fuzzy neural network model is put forward according to the MISO input model for the first time,and isolation hierarchical optimization algorithm is designed in the output layer and hidden layer.Secondly,the weighting valued parameters are optimized with the solving method of generalized inverse minimum norm and least square method for system of linear equations in output layer and hid-den layer.Finally,by MATLAB software for simulation experiments of MIMO polygonal fuzzy neural network.The results show that the network can improve the computing efficiency and convergent speed.In the third chapter:A new type polygonal Mamdnai fuzzy system is constructed by some polygonal fuzzy rules and the extension computation of polygonal fuzzy number at the first time,and a firefly optimization algoritlm for the weights parameters is designed based on fit-ness function,fluorcscein and decision radius.In addition,in the process of global searching by constantly looking for fireflies in the best,position to optimize the network model after a central connection weight parameters.Finally,through a simulation instance we verify the effectiveness of the firefly optimization algorithm.