文章基本信息

标题：Component-wise EM for Mixture Model Estimation
本地全文：下载
作者：Jun Sakuma ; Shin Ando ; Shigenobu Kobayashi 等
期刊名称：人工知能学会論文誌
印刷版ISSN：1346-0714
电子版ISSN：1346-8030
出版年度：2008
卷号：23
期号：3
页码：163-175
DOI：10.1527/tjsai.23.163
出版社：The Japanese Society for Artificial Intelligence
摘要：In the process of mixture model estimation using Expectation-Maximization (EM) methods, mixture densities are required to be measured at every step to obtain posterior probabilities. When the number of data n in a dataset or the number of mixtures m is large, the time complexity required for the evaluation of posterior probabilities is O(mn).
关键词：mixture model ; expectation maximization ; mean field theory ; Gaussian mixture ; clustering ; classification ; data mining