摘要
结合多样性密度和带负类的支持向量数据描述,提出了一种能够有效解决多示例问题的算法:MIL-NSVDD_DD。该算法首先通过多样性密度算法找出多示例问题中最优示例模型,然后通过使用带负类的支持向量数据描述对示例模型进行训练,以得到最终的分类器,用得到的分类器再对新包进行预测。最后通过实验表明了该算法的有效性。
In this paper,based on diverse density and support vector data description,we proposed MIL-NSVDD_DD algorithm which can solve multi-instance learning problem effectively.The algorithm firstly through diverse density method to find some optimal instance prototypes,secondly train these instance prototypes by Negative-SVDD can get a classifier,then use the classifier to predict new bag,finally the experimental results are promising.
出处
《工业控制计算机》
2012年第7期73-74,80,共3页
Industrial Control Computer
关键词
多示例学习
多样性密度
支持向量数据描述
机器学习
multiple-instance learning
diverse density
support vector data description
machine learning