Work

EM algorithm and its applications for mixture models

Year: 2012
Type: dissertation
Abstract: In statistics it is important to estimate the parameters of a data set if the family of the background distribution is known. The easiest way to do so is the well known Maximum Likelihood method. However, sometimes some part of the data is not known or the likelihood function cannot be maximized explicitly. In this case one needs to use iterative algorithms. One of the methods is the Expectation-Maximization (EM) algorithm. The EM algorithm, first published by Dempster, Rubin and Laird, is an iterative algorithm scheme that is applicable for several purposes in statistics and also in other sciences, like computational neuroscience. This is not a proper algorithm in terms of the computer science, since it does not describe the exact steps of the algorithm. Rather it is a scheme, though, we will refer to it as an algorithm. The implementations vary widely.
Language: English
Open Access status: green
OpenAlex ID: W7202346719
Cites:
Cited by:
Related to: 10