摘要
针对传统的近红外数据分析方法精度较低、,应用的局限性问题,本文提出了一种基于构造性算法的神经网络集成方法,由一个构造性算法决定个体网络中隐层节点的数量以保证个体网络的精确性,运用负相关学习算法和网络个体训练次数不同保证了网络个体的多样性。这种方法在近红外光谱分析中得到了成功的应用。
An ANN ensemble method based on contractive algorithm was established to overcome the low precision of traditional NIR spectroscopy and the limitation of application. In order to maintain accuracy among individual NNs,the number of hidden nodes in individual NNs is determined by a constructive approach. Negative correlation and different training epochs reflect the emphasis on diversity among individual. This method has been applied in NIR spectroscopy successfully.
出处
《光谱实验室》
CAS
CSCD
2005年第3期473-476,共4页
Chinese Journal of Spectroscopy Laboratory