摘要
参数选取问题一直是支持向量机研究的热点.虽然核校准(KTA)方法广泛应用于支持向量机参数优化问题中,但是它仍存在不足.以核矩阵为研究出发点,深入分析了采用核校准方法优化核参数对分类性能的影响,然后综合核校准方法和特征空间中样本集的分布提出了一种核校准改进方法.对比实验表明该算法是有效可行的.
The selection of parameter has been a significant issue for support vector machine. Although the Kernel Target Alignment （KTA） is widely used in the SVM parameters optimization problem, it has some drawbacks. Starting from kernel matrix, the influence on classification performance with the KTA method to optimize the kernel parameters is discussed. Then an improved KTA algorithm is presented by using the KTA and distribution of the sample set in feature space. The experiment results show that the improved KTA algorithm is effective and feasible.
出处
《烟台大学学报（自然科学与工程版）》
CAS
2013年第2期131-135,共5页
Journal of Yantai University(Natural Science and Engineering Edition)
基金
山东省自然科学基金资助项目(2009ZRB019CE)
关键词
参数优化
核校准
核矩阵
parameter optimization
kernel target alignment
kernel matrix
