论文部分内容阅读
讨论了BP小波神经网络在训练过程中减小误差函数时最优方向的确定和自适应调整学习率的方法。首先论证了小波神经网络的数学基础 ,然后讨论了BP小波神经网络的学习过程 ,重点讨论了减小误差函数最优方向的确定方法 ,即如何保证步长方向与负梯度方向一致 ,由此得出了自适应调整学习率的简便方法。该方法具有普遍性 ,有广泛的应用价值。仿真结果表明 ,采用最优梯度下降方向可以大幅度提高BP小波神经网络的学习速度。
This paper discussed how to determine the optimal direction and adjust the learning rate adaptively when BP neural network reduced the error function during the training process. First of all, the mathematical foundation of wavelet neural network is demonstrated. Then, the learning process of BP neural network is discussed, and the method to reduce the optimal direction of error function is discussed emphatically. How to ensure that the direction of step length is consistent with the negative gradient direction A simple way to adjust the learning rate adaptively. The method is universal and has a wide range of application value. Simulation results show that using the optimal gradient descent direction can greatly improve the learning speed of BP neural network.