【摘要】 本文提出了对ANF IS模糊神经网络的一种新的学习方法。由于梯度下降法和最小二乘法在训练ANF IS模糊神经网络时,尽管搜索局部极值的能力较强,但容易陷入局部极值,而遗传算法可以很快达到全局最优解附近。本文结合遗传神经网络、梯度下降法和最小二乘法各自的特点,提出一种混合算法,使得神经网络的收敛速度得到加快,精度得到提高。将该算法进行裂缝反演,取得了很好的反演效果。更多还原

【Abstract】 This thesis introduce a better study method for ANFIS neural network.. considering the character of the feature of GA neural networks which is good at global optimization and that of least square method and speedest descent method,we propose a new algorithm which combines GA with minimum sqare method and speedest descent method.By repeated practice on acoustic impedance inverse,The method can take a good effect on promoting the resolution.更多还原