论文部分内容阅读
基于ＬＭＳ的标准ＢＰ算法收敛速度极慢，而共轭梯度法要求精确的线性搜索，这在神经网络的高维权空间中是难以实现的。本文提出了一种新的ＢＰ学习算法，它采用一种对线性搜索要求不高的改进的共轭梯度法与一种简单的不精确线性搜索相结合，极大地提高了ＢＰ学习速度。经多次测试表明，与标准ＢＰ算法相比，该算法的效率提高了二个数量极。
The standard BP algorithm based on LMS has a very slow convergence rate, while the conjugate gradient method requires an accurate linear search, which is difficult to achieve in the high maintenance space of the neural network. In this paper, a new BP learning algorithm is proposed, which uses an improved conjugate gradient method which is less demanding on linear search and a simple imprecise linear search method, which greatly improves the learning speed of BP. After several tests show that compared with the standard BP algorithm, the efficiency of the algorithm increased two pole.