...
  • 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>International Journal of Innovative Computing Information and Control >INSTANCE-BASED K-NEAREST NEIGHBOR ALGORITHM FOR MULTI-INSTANCE MULTI-LABEL LEARNING
【24h】

INSTANCE-BASED K-NEAREST NEIGHBOR ALGORITHM FOR MULTI-INSTANCE MULTI-LABEL LEARNING

机译：基于实例的K-NEAREST NEIGHBOR算法用于多实例，多标签学习

获取原文
获取原文并翻译 | 示例
           
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Multi-instance multi-label learning (MIML) is proposed to tackle the problem represented by a bag of instances and associated with multiple labels which appears in a wide range of real-world tasks. Transforming the MIML problem into an equivalent problem is a very popular way to solve such learning work. However, such transformation may lose useful information encoded in training examples. In this paper we propose a new lazy learning algorithm - Instance-Based K-Nearest Neighbor (IB-KNN) which transforms MIML to multi-label learning (MLL), but makes full use of information between instances. Specifically, unlike most existing KNN methods for MIML problem which use the distance between bags, IB-KNN uses the distance between instances to discover the neighbor instances. Then', the neighbor instances vote to generate the preliminary results. After that, the problem becomes a standard MLL problem, and KNN method is used again to make the final prediction. We test the proposed algorithm on the COREL image data set and compare it with classical KNN based methods. The results show that IB-KNN achieves better performance.
机译：提出了多实例多标签学习（MIML），以解决由一系列实例代表的，与出现在现实世界中的各种任务中的多个标签相关联的问题。将MIML问题转化为等效问题是解决此类学习工作的一种非常流行的方法。但是，这样的转换可能会丢失训练示例中编码的有用信息。在本文中，我们提出了一种新的懒惰学习算法-基于实例的K最近邻居（IB-KNN），该算法将MIML转换为多标签学习（MLL），但充分利用了实例之间的信息。具体而言，与大多数现有的针对MIML问题的KNN方法使用包之间的距离不同，IB-KNN使用实例之间的距离来发现相邻实例。然后，邻居实例投票以生成初步结果。之后，该问题成为标准的MLL问题，并且再次使用KNN方法进行最终预测。我们在COREL图像数据集上测试了提出的算法，并将其与基于经典KNN的方法进行了比较。结果表明，IB-KNN具有更好的性能。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号