I am working on EM techniques to build clusters. Most of the websites and research papers have used Gaussian Mixture Model and EM techniques.
So, I need guidance related to probability distributions EM can use. And which one is best for performing clustering analysis using EM.
Regards,
It is not very clear what are your target distributions or goals of your mixure model building. Basically, the EM-algorithm can handle probability distributions as long as you formulate your model as a likelihood. A good general reference with many examples for mixture models is McLachlan & Peel (2000). "Finite mixture models". Also, McLachlan & Krishnan (1997). "The EM algorithm and extensions" is useful for learning the basics. Although it doesn't focus on mixture models, Skrondal & Rabe-Hesketh (2000) "Generalized latent variable modeling: multilevel, longitudinal and structural equation models" has a solid section on the EM algorithm.
Regards,
Timo
I agree with Timo Törmäkangas. In addition, I think we have to pay attention to the features of the data. A trade-off between the capability of the family of distributions and its complexity is important.
See this link:
https://machinelearningmastery.com/expectation-maximization-em-algorithm/
- Badges
- Science topic