基于人工神经网络的改进共轭梯度算法研究

Improved conjugate gradient algorithm based on ANN

针对前馈式神经网络结构,提出用一种改进共轭梯度算法建模的最速下降搜索迭代的新方法,并把它用于神经网络。同时用神经网络来预报经济指标,通过预测中使用的基本模型和训练算法,提出了单因素非线性自回归和多因素非线性回归两种神经网络预报模型,并对四川省的社会总产值进行预测,结果表明此神经网络用于经济预测是一种新的、更精确、更有效的预测方法。

In accordance with the structure of feed-forward neural network, a new method of iteration of improved conjugated gradient algorthm is presented. The index of economic can forecasted with neural network. By both the basic model and the train algorithm of forecastion, the model of asingle factor non-line regression-self and multi-fotor non-line are put forward and the social general production in Sichuan province is forecasted. The result show forecasting the ecomomy with neural network is new, much accuracy and effective method.