Fast Real-time Decision Approach of Support Vector Data Description
-
摘要: 为了提高一类支持向量数据描述(Support vector data description, SVDD)对未知样本的决策速度,本文从样本的核特征空间出发, 利用核超球球心在原始样本特征空间中的原像,提出一种SVDD的快速决策方法(Fast decision approach of SVDD, FDA-SVDD),使得SVDD的决策复杂度从O(n)降低到O(1). 同时,对球心原像所在空间进行了分析,并在此基础上给出了两种原像逼近方法.多种真实数据集实验表明, FDA-SVDD方法在保证测试精度的同时,能快速实现对未知样本的决策.Abstract: For improving the decision speed of one-class SVDD, a fast decision approach is proposed in this paper, called FDA-SVDD, by utilizing the preimage in original feature space corresponding to the center of sphere in kernel feature space, by which the decision complexity of SVDD is reduced from O(n) to O(1). Meanwhile, two approximate algorithms for finding preimage are presented based on analyzing its position in the original space. Experimental results on extensive datasets show that the proposed method can not only guarantee testing accuracy but also fast classify unknown samples.
-
点击查看大图
计量
- 文章访问数: 1868
- HTML全文浏览量: 54
- PDF下载量: 1181
- 被引次数: 0
下载:
