针对前向神经网络现有BP学习算法的不足,结合非线性最优化方法,提出一种基于拟牛顿法的神经元网络学习算法。该算法有效地改进了神经元网络的学习收敛速度,取得了比常规BP算法更好的收敛性能和学习速度。 Aiming at the insufficiency of existing BP learning algorithm in feedforward neural network, a nonlinear learning algorithm based on quasi-Newton algorithm is proposed based on nonlinear optimization method. The algorithm effectively improves the learning convergence speed of neural networks and achieves better convergence performance and learning speed than the conventional BP algorithm.