摘要
文章研究了基于遗传算法的神经网络油水层识别方法,针对神经计算存在因输入信息空间维数较大而使网络结构复杂、训练时间长,以及因冗余属性使网络拟合精度不高等缺点,提出了基于粗集属性约简方法降低了输入信息的空间维数、减少了运算量和简化了神经网络的拓扑结构,利用遗传算法提高神经网络的训练速度。实验结果表明:将混合智能计算方法应用于油水层识别中效果显著,其学习训练速度和拟合精度远优于传统BP神经网络算法。
In this paper,a composition intelligence computing method is suggested for an oil-water layer recognition.The redundant condition attributes are reduced based on rough set attribute simplification algorithm so that an oil-water layer neural network recognition system can be simplified to improve network composition.An optimization computation speed of neural network is improved by a BP learning with a genetic algorithm.Simulation result shows that the effect in oil-water layer recognition is improved by the composition intelligence computing method proposed here.
出处
《四川理工学院学报（自然科学版）》
CAS
2010年第5期590-593,共4页
Journal of Sichuan University of Science & Engineering(Natural Science Edition)
基金
国家高技术研究发展计划(863计划)第四批课题(2008AA11A134)
关键词
属性约简
神经网络
遗传算法
油水层识别
attribute reduction
neural network
genetic algorithm
oil-water layer recognition
