摘要
本文提出一种运用人工神经网络结合正交变换的方法，即通过正交变换滤除噪声，通过交叉验证确定网络最佳构型，以充分发挥正交变换和神经网络各自的长处，避免出现过拟合，实现更准确的预报。作为一个应用实例，对初轧钢板坯样本集进行了处理并预报了不同工艺参数下的钢坯废品率。结果表明，用神经网络结合正交变换新方法可达到很好的预报效果。
In this paper we proposed a new algorithm of artificial neural network with orthogonal expansion. The orthogonal expansion was used to eliminate the noise in data set and the cross-validation was used to determine the optimal structure of neural network.Since the algorithm combines the advantage of orthogonal expansion and neural network, the oveffitting problem can be avoided and the exact prediction can be obtained. As an example,the prediction of waste rate of first-rolled steel slab was discussed and the results indicated that a good precision of prediction was given by the new method.
出处
《计算机与应用化学》
CAS
CSCD
1997年第2期127-132,共6页
Computers and Applied Chemistry
基金
国家自然科学基金
关键词
人工神经网络
正交变换
初轧
钢板坯
Artificial neural network,Orthogonal expansion, First-rolled steel slab
