Improved initialization of the EM algorithm for mixture model parameter estimation

Panić, Branislav (author), Klemenc, Jernej (author), Nagode, Marko (author)

A commonly used tool for estimating the parameters of a mixture model is the Expectation-Maximization (EM) algorithm, which is an iterative procedure that can serve as a maximum-likelihood estimator. The EM algorithm has well-documented drawbacks, such as the need for good initial values and the possibility of being trapped in local optima. Nevertheless, because of its appealing properties, EM plays an important role in estimating the parameters of mixture models. To overcome these initialization problems with EM, in this paper, we propose the Rough-Enhanced-Bayes mixture estimation (REBMIX) algorithm as a more effective initialization algorithm. Three different strategies are derived for dealing with the unknown number of components in the mixture model. These strategies are thoroughly tested on artificial datasets, density-estimation datasets and image-segmentation problems and compared with state-of-the-art initialization methods for the EM. Our proposal shows promising results in terms of clustering and density-estimation performance as well as in terms of computational efficiency. All the improvements are implemented in the rebmix R package.

• Language: English
• Keywords: mixture model, parameter estimation, EM algorithm, REBMIX algorithm, density estimation, clustering, image segmentation
• Work type: Article
• Typology: 1.01 - Original Scientific Article
• Organization: FS - Faculty of Mechanical Engineering
• Publication status: Published
• Publication version: Version of Record
• Year: 2020
• Number of pages: 29 str.
• Numbering: Vol. 8, iss. 3, art. 373
• PID: 20.500.12556/RUL-114907
• UDC: 519.254(045)
• ISSN on article: 2227-7390
• DOI: 10.3390/math8030373
• COBISS.SI-ID: 17112347
• Publication date in RUL: 30.03.2020

Record is a part of a journal

• Title: Mathematics
• Shortened title: Mathematics
• Publisher: MDPI AG
• ISSN: 2227-7390
• COBISS.SI-ID: 523267865

Licences

• License: CC BY 4.0, Creative Commons Attribution 4.0 International
• Link: http://creativecommons.org/licenses/by/4.0/
• Description: This is the standard Creative Commons license that gives others maximum freedom to do what they want with the work as long as they credit the author.
• Licensing start date: 07.03.2020

Secondary language

• Language: Slovenian
• Keywords: mešani model, ocena parametrov, EM algoritem, REBMIX algoritem, ocena gostote, porazdelitev verjetnosti, grozdenje, segmentacija slik

Projects

• Funder: ARRS - Slovenian Research Agency
• Project number: 1000-18-0510