论文部分内容阅读
本文提出一种规则结论部分的语言变量具有离散隶属度函数的、基于Mamdani形规则的新神经模糊系统,并描述了它的学习算法。新神经模糊系统由模糊推理系统及其一对应的神经网络系统构成。在只有训练数据的情况下,首先提出了一种基于RBF神经网络的模糊建模方法。而在模糊推理系统由模糊建模或者直接由专家经验知识确定后,应用梯度下降法优化神经网络系统参数。倒立摆控制和时间序列预测的仿真试验体现了本文提出的新的神经模糊系统的可用性和优越性。
In this paper, a new neural fuzzy system based on the Mamdani-type rule whose language variables have discrete membership functions is proposed and its learning algorithm is described. The new neuro-fuzzy system consists of a fuzzy inference system and a corresponding neural network system. In the case of training data only, a fuzzy modeling method based on RBF neural network is first proposed. However, when the fuzzy inference system is determined by fuzzy modeling or directly by the expert experience, the gradient descent method is used to optimize the parameters of the neural network system. The simulation experiments of inverted pendulum control and time series prediction show the usability and superiority of the new neuro-fuzzy system proposed in this paper.