CAJ | 학술논문

提出一种改进的模糊神经网络混合学习算法，运用遗传算法优化构成隶属函数的网络结构，运用最小二乘法进行解模糊。具有更高的学习精度和更快的收敛速度，解决了在多变量系统中采用模糊神经网络时学习收敛慢且易陷入局部极小点的问题。
제출일충개진적모호신경망락혼합학습산법，운용유전산법우화구성대속함수적망락결구，운용최소이승법진행해모호。구유경고적학습정도화경쾌적수렴속도，해결료재다변량계통중채용모호신경망락시학습수렴만차역함입국부겁소점적문제。
An Improved hybrid algorithm for fuzzy neural network (FNN) is presented. Genetic algorithm and Least-Square technique are introduced in making the network structure for the membership function and doing the defuzzification, respectively. The new algorithm has more accurate precision and faster convergent speed. When the multi-variable system is controlled using FNN, the learning process is always slowly and traps in the local convergence. These problems have been solved.