• 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 >A NOVEL PROTOTYPE REDUCTION APPROACH FOR SUPERVISED LEARNING
【24h】

A NOVEL PROTOTYPE REDUCTION APPROACH FOR SUPERVISED LEARNING

机译：监督学习的新型原型减少方法

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

We propose a similarity-based prototype reduction algorithm to reduce the training set size for supervised learning. Training patterns are input to the algorithm one by one and grouped into blobs through similarity tests. The statistical mean of each blob is regarded as a prototype representing all the patterns included in the blob. The collection of such means can then be used to substitute the original training set, and, consequently, the training set for later supervised learning is reduced. This approach has several advantages. The distribution of the data contained in each blob is statistically well described. Each obtained prototype is a good representative of the patterns included in the corresponding blob. Different numbers of representatives are extracted automatically according to the similarity relationship among and the distribution of the original training patterns. Furthermore, our method can be applied efficiently to both regression and classification problems. Experimental results show that the proposed method performs more effectively than other prototype reduction methods.
机译：我们提出了一种基于相似度的原型约简算法，以减少监督学习的训练集大小。训练模式被一一输入到算法，并通过相似性测试分组为斑点。每个斑点的统计平均值被视为代表斑点中包括的所有模式的原型。然后可以使用此类工具的集合来替代原始训练集，因此减少了用于以后监督学习的训练集。这种方法有几个优点。统计上很好地描述了每个斑点中包含的数据分布。每个获得的原型都很好地代表了相应斑点中包含的模式。根据原始训练模式之间的相似关系和分布，自动提取不同数量的代表。此外，我们的方法可以有效地应用于回归和分类问题。实验结果表明，该方法比其他原型约简方法更有效。

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