In multi-instance learning, the training set is composed of labeled bags each consists of many unlabeled instances, and the goal is to learn some classifier from the training set for correctly labeling unseen bags. In the past, some research about multi-instance learning work through adapting single-instance learning algorithms to the multi-instance representation, others try to propose some new methods to find the relationship between instances and bags and use the result to solve the problem. In this paper, a new algorithm-Concept mining algorithm is proposed. This algorithm starts from adapting the representation of the bag and re-represent a bag as a d dimensional vector- Concept vector of bag with the R-pattern discovery which is a method using in document filter. Because after re-representing the data set is not "multi" again, some propositional single-instance learning algorithms can be used to solve multi-instance learning problem effetely.%在多示例学习问题中,训练数据集里面的每一个带标记的样本都是由多个示例组成的包,其最终目的是利用这一数据集去训练一个分类器,使得可以利用该分类器去预测还没有被标记的包.在以往的关于多示例学习问题的研究中,有的是通过修改现有的单示例学习算法来迎合多示例的需要,有的则是通过提出新的方法来挖掘示例与包之间的关系并利用挖掘的结果来解决问题.以改变包的表现形式为出发点,提出了一个解决多示例学习问题的算法——概念挖掘算法.该算法利用原本用于文本过滤的R-模式发现法将包表示成一个d维向量——概念向量.经过重新表示后,原有的多示例数据集已经不再“多示例”,以至于一些现有的单示例学习算法能够被用来高效地解决多示例学习问题.
展开▼
