Type-2 Fuzzy Systems Theory And Its Optimization Research

After more than 60 years of development,artificial intelligence is shocking human beings with more and more powerful kinetic energy.In particular,artificial intelligence based on deep neural network technology has achieved many outstanding results by relying on its powerful capabilities,but it is not an intelligence with the ability to"analogical association" after all.Inspired by human perception behavior,Zadeh proposed the theory of perceptual computing,which uses fuzzy theory to realize the utilization of knowledge and data,and provides a solution for the next generation of intelligence with strong interpretability and robustness.Fuzzy theory was initially researched around type-1 fuzzy systems,but in order to enhance its uncertainty description ability,Zadeh introduced type-2 fuzzy systems.Type-2 fuzzy system can describe ambiguity,inaccuracy and controversy more accurately by using three-dimensional structure,and can construct uncertainty within and between individuals at the same time,but its structure complexity is greatly increased,especially in the output processing part.The type-reduction seriously restricts the efficiency of the system and greatly hinders the practical application.In order to enhance the ability of fuzzy systems to deal with high-dimensional nonlinear systems,the hierarchical fuzzy structure is extended to type-2 fuzzy systems,so that type-2 hierarchical fuzzy systems have both deep reasoning ability and strong uncertainty description ability,and at the same time solve the disaster of dimensionality.However,the structure of the type-2 hierarchical fuzzy system is more complex.How to reflect the combination of knowledge and data in the type-2 hierarchical fuzzy system and realize its parameter optimization is particularly important.On the basis of analyzing the existing problems of the current type-2 fuzzy systems,this article take a deep dive into the efficient type-reduction of the interval type-2 fuzzy system,the efficient type-reduction of the general type-2 fuzzy system and the parameter optimization of the interval type-2 hierarchical fuzzy system,the main contents are:1.Aiming at the problem of efficient type-reduction for interval type-2 fuzzy systems,an improved enhanced opposite search algorithm is proposed.The algorithm is based on the location properties of the endpoint values of the type-reduction set,and the one-way iterative search algorithm is obtained by expressing the upper membership value as the sum of the lower membership and the membership width,and using the best initial value,it can maximize the reduce the number of iterative searches and improve computational efficiency.This paper proves the global convergence of the algorithm;analyzes the computational complexity of the algorithm,and verifies that the algorithm has fewer iterations and faster computation time.Finally,the simulation experiment verifies the high efficiency of the reducer.2.Aiming at the problem of efficient type-reduction for general type-2 fuzzy systems,a method for calculating the centroid based on the horizontal plane representation is proposed.The algorithm decomposes the general type-2 fuzzy set into horizontal planes with different height values,and calculates the centroid for each plane from top to bottom according to the inclusive relationship of the end points of the plane centroid intervals of different heights;At the same time,in the adj acent planes,the switch point of the previous plane is used as the initial value of the calculation of the centroid of the next plane,thereby further reducing the number of iterations and calculation time required by the algorithm.Finally,the efficiency of the algorithm is verified by simulation experiments.3.Aiming at the parameter optimization problem of interval type-2 hierarchical fuzzy system,a two-stage optimization algorithm based on gradient and swarm intelligence mixed intelligence is proposed.The algorithm first constructs a type-1 hierarchical fuzzy system,and optimizes its parameters by using the gradient algorithm based on the combination of mini-batch gradient,regularization and AdaBound technology;Then,it is converted into an interval type-2 hierarchical fuzzy system,and then the particle swarm algorithm with adaptive weight is used to optimize the fuzzification parameter interval,so as to obtain the optimal interval type-2 hierarchical fuzzy system.Finally,the effectiveness of the algorithm is verified by simulation experiments.