摘要
提出了一种新的约束信息扩展方法。该方法先利用给定的标记信息建立凸壳,然后扫描整个数据集,选择在凸壳内的数据点作为候选集并做进一步判断;对于凸壳间的公共数据,采用凸多边形最优三角剖分的方法来确定这些数据最终应加入的标记信息集。该约束信息扩展方法在四类数据集上验证了算法的有效性。
This paper proposes a new way of constraints extending. Use the given constraints to structure convex hulls. And then scan the data set, choose the samples which are in or on the convex hulls for the next judgment. To those samples which are public between different hulls, use the optimal triangulations algorithm to determine the final constraint set they should belong to. The accuracy rate is verified in four types of data samples.
出处
《计算机工程与应用》
CSCD
2014年第4期173-176,共4页
Computer Engineering and Applications
基金
国家自然科学基金(No.61103129)
江苏省科技支撑计划资助项目(No.BE2009009)
关键词
半监督
凸壳
约束信息扩展
最优三角剖分
semi-supervised
convex hulls
constraints extending
optimal triangulations
