Earticle

본 논문에서는 Mackey-Glass시계열의 예측에서 유전자알고리즘을 이용하는 구조적인 동정과 뉴로퍼지에 의한 파라미터 동정의 학습방법과 하이브리드 시스템을 제안하였다. 본 방법은 두 가지로 구성되었다. 하나는 입력공간에 대한 분할을 통하여 유전 알고리즘을 이용하여 퍼지 규칙베이스를 구축하고 다른 하나는 이 규칙베이스를 토대로 기울기 최하강법을 이용하여 제어규칙의 변수에 대한 파라미터 동정이다. 제안된 방법을 성능을 검증하기 위하여 입력의 패턴을 시간간격에 따라서 x(t-3), x(t-6)과 x(t-9)의 세 가지로 구성하였다. 많은 시뮬레이션을 통하여 유전알고리즘에 의한 구조적인 동정으로 인하여 학습초기에 오차가 작은 것을 알 수 있었다. 표2에서와 같이 성능을 확인 할 수 있었다.

In this paper, we introduce the evolutionary design methodology of fuzzy inference systems by means of fuzzy partition of input space. The rules of the proposed fuzzy model are realized with the aid of the fuzzy partition of input space generated by fuzzy c-means clustering algorithm. The number of the partition of input space is equal to the number of clusters. And the individual partitioned spaces describe the fuzzy rules. Due to these characteristics, we may alleviate the problem of the curse of dimensionality. The consequence part of the rule is represented by polynomial functions. We also consider the evolutionary optimization to determine the structure and estimate the values of the parameters of the model using real-coded genetic algorithms. Numerical examples are included to evaluate the performance of the proposed model.

In order to optimize fuzzy modeling of nonlinear system, we proposed a optimal fuzzy model according to the characteristic of I/O relationship, HCM method, the genetic algorithm, and the objective function with weighting factor. A conventional fuzzy model has difficulty in definition of membership function. In order to solve its problem, the premise structure of the proposed fuzzy model is selected by both the partition of input space and the analysis of input-output relationship using the clustering algorithm. The premise parameters of the fuzzy model are optimized respectively by the genetic algorithm and the consequence parameters of the fuzzy model are identified by the standard least square method. Also, the objective function with weighting factor is proposed to achieve a balance between the performance results for the training and testing data.

In this paper, we present an approach to the structure identification based on the input space partition methods and to the parameter identification by hybrid learning method in neuro-fuzzy system. The structure identification can automatically estimate the number of membership function and fuzzy rule using grid partition, tree partition, scatter partition from numerical input-output data. And then the parameter identification is carried out by the hybrid learning scheme using back-propagation and least squares estimate. Finally, we sill show its usefulness for neuro-fuzzy modeling to truck backer-upper control.