【摘要】 针对小波神经网络常用的误差反传算法存在着易陷入局部极小点和对初值参数要求较高的缺点,结合遗传算法自适应全局优化搜索能力与小波神经网络良好的时频局部特性,提出了一种有效的学习训练途径。该方法首先应用遗传算法确定网络的初始参数,然后转入纯小波神经网络进行训练,大大加快了网络的收敛速度。网络训练时采用共轭梯度学习算法并对此算法进行了改进,有效的克服了梯度学习算法容易陷入局部极小的缺点。通过二阶倒立摆的控制仿真和实物控制,验证了算法的有效性。更多还原

【Abstract】 The error back-propagation algorithm of wavelet neural network has some weakness that it is easily trapped into local minimum point and that it is hard to determine its initial easy local minimum point and the initial values of parameter.To overcome these shortcomings an effective approach is proposed by combining the global searching ability of genetic algorithm and the good local performance in both time and frequency fields of the wavelet network.In this method,the initial values of parameters of the network are firstly determined by genetic algorithm,then the wavelet neural network is trained,thus its convergence is speeded up greatly.Moreover,an improved conjugate gradient algorithm is used to train the network and to overcome the shortcoming of easily trapping into local minimum points.Finally,the proposed method is simulated and realized in the double inverted pendulum control system,and its effectiveness is demonstrated.更多还原