为提高神经网络的训练速度,提出一种自适应确定带动量项BP算法中动量因子的方法.在学习率为常数情况下,根据误差函数关于权值向量的梯度变化情况,自适应调节动量因子.数值试验表明,该方法对离线和在线训练均有效,且在收敛速度和算法稳定性上优于常动量因子的BP算法. In order to improve the training speed of neural network, a method to adaptively determine the momentum factor in the BP algorithm with momentum is proposed.When the learning rate is constant, according to the gradient function of the error function about the weight vector, the momentum factor Numerical experiments show that this method is effective for both offline and online training, and is superior to the normal momentum factor in convergence rate and algorithm stability.