针对传统的ＢＰ或ＧＡ对模糊神经网络的识别应用存在收敛容易陷入局部极小 识别率低下等问题 提出一

种基于ＢＦＧＳ的混合遗传算法 其基本思想为 首先构造一种前馈型模糊神经网络结构 然后用遗传算法进化若干代

后 当目标函数的梯度或者范数小于预先设定值 则改用ＢＦＧＳ算法进行优化识别 仿真实验表明 对比ＧＡ该算法

收敛速度较快 识别精度提高了约７％ 能够较好地应用于一类模糊神经网络的识别-In traditional BP or GA to identify the application of fuzzy neural network in convergence of easily falling into local minimum problem of low recognition rate is proposed



A hybrid genetic algorithm based on BFGS and its basic idea is first to construct a feedforward fuzzy and genetic algorithm is used to evolve neural network structure for several generations



When the gradient of the objective function or norm less than the preset value is used to optimize BFGS algorithm recognition experiment compared the algorithm GA



The recognition accuracy of fast convergence speed is improved about 7　recognition can be applied to a class of fuzzy neural networks