摘要
提出了一种基于支持向量机回归的去噪方法。传统的统计学理论的去噪方法由于训练样本数的限制,在实际应用中普遍存在推广能力不足的问题。本文借助支持向量机在小样本情况下良好的推广能力,利用支持向量机回归逼近数据去除噪声。使用该方法对大地电磁测深视电阻率进行了处理。实际资料处理结果表明,基于支持向量机回归的去噪方法,较好地消除了噪声的影响,同时对消除部分测点的地形与局部异常体的影响也有一定的作用。
In this paper, we present a method for eliminating noise based on support vector machine regression. Owing to the limit samples in actual application, some noise reduction methods based on traditional statistical learning theory are not ideal. With the help of the good generalization of support vector machine in small samples, this approach depicts the signal and noise with support vector machine. The processing results of magnetotelluric sounding data by the method show that effect of noise is reduced.
出处
《工程地球物理学报》
2005年第3期191-194,共4页
Chinese Journal of Engineering Geophysics
基金
湖北省自然科学基金项目(项目编号:2004ABA043)
国家自然科学基金项目(项目编号:40174040)。
关键词
支持向量机
去噪
大地电磁测深
函数回归
support vector machine
noise elimination
magnetotelluric sounding
function regression
