摘要
由于支持向量机对样本中的噪声及孤立点非常敏感,因而在解决非线性、高维数、不确定问题时,使用模糊支持向量机比使用支持向量机的效果要好。在模糊支持向量机中,模糊隶属度函数的建立是关键也是难点。一般,模糊隶属度是在原始空间中根据样本点的相互距离及到类中心的距离创建的。考虑样本间的密切度,在特征空间中利用混合核函数建立一种新的模糊隶属度。通过试验比较多项式核函数、高斯径向基核函数与混合核函数,可看出新方法表现出了它的优越性。
Support vector machine is sensitive to the noises and outliers in the training samples, so fuzzy support vector machine precede support vector machine in solving the problem of non - linearity, high dimension and uncertainty. The choice of fuzzy membership is the key and difficulty for fuzzy support vector. Generally, the fuzzy membership is established according to the distance of between sample points and its cluster center. A new fuzzy membership function is established, considering the relation among samples using mixed kernel function,based on mixed kernel function. The experiments show that fuzzy support vector machine with the new fuzzy membership is superior through comparing mixed kernel function with Polynomial kernel function and Gaussian RBF kernel function.
出处
《计算机技术与发展》
2010年第2期9-11,15,共4页
Computer Technology and Development
基金
国家自然科学基金项目(10371106
10471114)
江苏省高校自然科学基金项目(04KJB110097
08KJB520003)
南京邮电大学攀登计划(NY207064)
关键词
模糊支持向量机
模糊隶属度
混合核函数
fuzzy support vector machine
fuzzy membership
mixed kernel function
