...
  • 主题
高级搜索  >
历史搜索 清空历史
首页> 外文期刊>Neural processing letters >A Dynamic Programming Framework for Large-Scale Online Clustering on Graphs
【24h】

A Dynamic Programming Framework for Large-Scale Online Clustering on Graphs

机译：图形上的大型在线聚类的动态编程框架

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

As a fundamental technique for data analysis, graph clustering grouping graph data into clusters has attracted great attentions in recent years. In this paper, we present DPOCG, a dynamic programming framework for large-scale online clustering on graphs, which improves the scalability of a wide range of graph clustering algorithms. Specifically, DPOCG first identifies the nodes whose states are unchanged compared with the states at the previous time on a large-scale graph, then constructs these unchanged nodes as supern-odes, which greatly reduces the size of the graph at the current time, and collapses nodes whose degrees are less than a predefined threshold. Based on our density-based graph clustering algorithm (DGCM), DPOCG partitions the reduced graph into clusters. In addition, we theoretically analyze DPOCG in terms of supernode generation, clustering on reduced graph, and computational complexity. We evaluate DPOCG on a synthetic dataset and seven real-world datasets, respectively, and the experimental results show that DPOCG consumes less running time and improves the efficiency of clustering.
机译：作为数据分析的基本技术，近年来，图形聚类为集群将图形数据进行了吸引了巨大的关注。在本文中，我们展示了DPOCG，是在图形上进行大规模在线聚类的动态编程框架，这提高了各种图形聚类算法的可扩展性。具体地，DPocg首先识别与在大规模图中上一段时间上的上次在前时间的状态相比的节点，然后将这些不变的节点构造为叠加杂物，这大大减少了当前时间的曲线的大小，并且折叠度度小于预定阈值的节点。基于我们基于密度的图形聚类算法（DGCM），DPOCG将缩小的图分成群集。此外，我们在理论上，从大型码发电，在减小的图表上聚类和计算复杂性方面地分析DPOCG。我们分别评估DPOCG和七个现实世界数据集，实验结果表明，DPOCG消耗了更少的运行时间并提高了聚类的效率。

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