论文部分内容阅读
首次采用遗传算法（ＧＡ）设计模糊小脑模型神经网络（ＦｕｚｚｙＣＭＡＣ）的隶属函数．提出一个自适应ＧＡ优化算法，并且以优化模糊小脑模型ＦｕｚｙＣＭＡＣ学习正弦曲线．仿真实例表明，采用自适应ＧＡ方法优化的ＦｕｚｙＣＭＡＣ学习精度比标准小脑模型ＣＭＡＣ提高大约三个数量级、比标准ＦｕｚｚｙＣＭＡＣ（三角形隶属函数）提高一个数量级．自适应ＧＡ方法优化的ＦｕｚｙＣＭＡＣ学习速度比普通ＧＡ优化的速度快且进化过程的振荡明显减小，仿真证明该方法比普通ＧＡ优化方法稳定，收敛效果好．
For the first time, genetic algorithm (GA) was used to design the membership function of FuzzyCMAC. An adaptive GA optimization algorithm is proposed, and a sinusoidal curve is learned by the FuzyCMAC algorithm. Simulation results show that the learning accuracy of FuzyCMAC optimized by adaptive GA method is about three orders of magnitude higher than that of standard cerebellar model CMAC, which is an order of magnitude higher than that of standard FuzzyCMAC (triangle membership function). The learning speed of FuzyCMAC optimized by adaptive GA is faster than that of ordinary GA and the oscillation of evolution is obviously reduced. Simulation results show that this method is more stable than conventional GA and has a good convergence effect.