Hierarchical representations of network data with optimal distortion bounds

Hierarchical representations of network data with optimal distortion bounds
复制标题

具有最佳失真范围的网络数据的分层表示

DOI:
10.1109/acssc.2016.7869701
复制
发表时间:
2016
期刊:
2016 50th Asilomar Conference on Signals, Systems and Computers
影响因子:
--
通讯作者:
F. Mémoli
F. Mémoli
中科院分区:
--
文献类型:
--
作者:
Zane T. Smith;Samir Chowdhury;F. Mémoli

文献摘要

被引文献

相似文献

单链层次聚类是无监督学习中的一种工具,它已经在有限度量空间中得到了充分的刻画,但不适用于一般网络的无限制设置。我们遵循最近的一系列工作来完成对一般网络的刻画,此外,我们还提供了将我们的方法应用于网络数据时丢失多少信息的量化界限。即使在有限度量空间的情况下,这些界也是新的。最后,我们提出了一种称为树图的结构,它提供了将我们的方法应用于网络数据集的结果的可视化总结。
Single linkage hierarchical clustering is a tool in unsupervised learning which has been fully characterized for finite metric spaces, but not for the unrestricted setting of general networks. We follow a recent line of work to complete the characterization for general networks, and moreover, we provide quantitative bounds on how much information is lost when applying our method to network data. These bounds are novel even in the setting of finite metric spaces. Finally, we propose a construction called a treegram that provides a visual summary of the result of applying our method to a network data set.