An investigation of computational and informational limits in gaussian mixture clustering.pdf

下载文档

2
0
约3.78万字
约 9页
2017-04-13 发布于江苏
举报
版权申诉
保障服务

An investigation of computational and informational limits in gaussian mixture clustering.pdf

1、本文档共9页，可阅读全部内容。
2、有哪些信誉好的足球投注网站（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。
3、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击这里二次下载。
4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。

An investigation of computational and informational limits in gaussian mixture clustering

Department of Computer Science 6 King’s College Rd, Toronto University of Toronto M5S 3G4, Canada fax: +1 416 978 1455 February 10, 2006 UTML TR 2006–002 An Investigation of Computational and Informational Limits in Gaussian Mixture Clustering Nathan Srebro Department of Computer Science, University of Toronto Gregory Shakhnarovich Department of Computer Science, Brown University Sam Roweis Department of Computer Science, University of Toronto Abstract We investigate under what conditions clustering by learning a mixture of spherical Gaussians is (a) computationally tractable; and (b) statistically pos- sible. We show that using principal component projection greatly aids in recovering the clustering using EM; present empirical evidence that even us- ing such a projection, there is still a large gap between the number of samples needed to recover the clustering using EM, and the number of samples needed without computational restrictions; and characterize the regime in which such a gap exists. An Investigation of Computational and Informational Limits in Gaussian Mixture Clustering Nathan Srebro nati@ Department of Computer Science, University of Toronto Gregory Shkhnarovich gregory@ Department of Computer Science, Brown University Sam Roweis roweis@ Department of Computer Science, University of Toronto Abstract We investigate under what conditions clus- tering by learning a mixture of spherical Gaussians is (a) computationally tractable; and (b) statistically possible. We show that using principal component projection greatly aids in recovering the clustering using EM; present empirical evidence that even using such a projection, there is still a large gap between the number of samples needed to re- cover the clustering using EM, and the num- ber of samples needed without computational restrictions; and characterize the regime in which such a gap exists. 1. Introduction Consider clustering a collection of points by fitting a mixture-of-Gaussians model to the data.