摘 要:本文研究的是一种神经网络与共轭梯度算法相结合的改进算法。首先分别介绍了共轭梯度算法和神经网络的相关概念和知识,然后提出了一种新的改进的算法,即将具有先验知识的神经网络引入到共轭梯度法中,最后通过两个实例进行计算并且与普通的共轭梯度法进行对比得出以下结论:改进后的算法的稳定性好,而且还具备共轭梯度法本身的长处,收敛速度快。38904 毕业论文关键词:先验知识;共轭梯度法;神经网络
Research on the Improved Conjugate Gradient Method Based on
the Neural Network
Abstract: The paper studies an improved algorithm combined the neural network with conjugate gradient algorithm. Firstly, the related concepts and knowledge about the conjugate gradient algorithm and neural network are introduced. Secondly, a new improved method is proposed. That is the neural network with a prior knowledge is leaded up to the conjugate gradient method. Finally, through the analysis of two examples and comparison with the common conjugate gradient method, the proposed algorithm has good stability, fast convergence and the advantages which the conjugate gradient method possesses by itself.
Key words: A prior knowledge; Conjugate gradient method; Neural network
目 录