摘要
BP神经网络具有很强的非线性函数逼近能力。通过Matlab编程实现了BP网络对一个二元函数的逼近,针对BP网络学习速度慢的缺点,采用含动量项的学习算法提高了收敛速度。通过Matlab仿真方法,研究了学习率和动量因子对算法学习速度的影响。
BP neural network has strong ability of approaching nonlinear function. Matlab programming of BP network was implemented to approach a 2-dimentional function. The convergence speed was expedited via the momentum in the learning algorithm. The influence of learning rate and momentum against the learning speed were discussed.
出处
《浙江交通职业技术学院学报》
CAS
2007年第4期45-48,共4页
Journal of Zhejiang Institute of Communications
