BP网络的BP-G-S学习算法

The BP G S Learning Algorithms of BP Network

为了解决ＢＰ网络在学习过程中存在收敛慢的缺点，文中将ＧａｕｓｓＳｅｉｄｅｌ 迭代法的基本思想与ＢＰ算法结合，将每一个最新修正的权值反映到下一个权值的修正中，提出了一种新的ＢＰＧＳ学习算法来加速ＢＰ网络的收敛。文章最后的计算机仿真说明ＢＰＧＳ总体上可以减少学习的时间，尤其当误差值逼近最小点时效果明显。

To solve the problem of the slowness of convergence in the learning process of BP network,this paper integrates the idea of Gauss Seidel algorithms into BP algorithms and makes use of the newly corrected weight while correcting next weight.The imitation of computer proves that BP G S algorithm may decrease learning time in the whole.The effect is obvious especially when the error value is near to the optimal point.