摘要
针对一般BP算法收敛速度慢和易陷入局部极小值的缺陷,提出利用一种自适应学习速率动量梯度下降反向传播算法对网络进行训练。该算法使BP神经网络学习速率和稳定性得到提高,并将这种改进的BP网络应用于个人信用评估系统,验证了该方法的正确性和有效性。
A reverse transmission calculation of self adaptive learning rate with momentum gradient reduction is described in the paper against the problems of slow convergence and easy trapping into smallest spot.The calculation has made the BP net upgrade the rate and stability of learning. BP and the improvement of the network used a personal credit assessment system to verify the correctness of the method and effectiveness.
出处
《齐齐哈尔大学学报（自然科学版）》
2008年第5期15-18,共4页
Journal of Qiqihar University(Natural Science Edition)
关键词
BP算法
函数梯度
神经网络
BP algorithm
function grads
neural network
