EM algorithm for Mixture models (test version)

Usage

1. You can select Gaussian Mixture Fitting by choosing "GaussMix", or Multiple Line Fitting by choosing "LineMix".
2. Click in a image plane to add sample points.
3. You can also add unformly distributed sample points by clicking "RandomPts".
4. "ClearPts" enables to clear all sample points
5. You can select the number of kernels by choosing 1 to 5.
6. You can initialize the kernels by clicking "InitKernels".
7. You can start EM algorithm by selecting "EM Run" or "EM 1 Step"(one step)
8. You can stop "EM Run" by selecting "EM Stop".

Bugs
If Gaussians or Lines disappear by EM, initialize kernels by click "InitKernels".

Source (created in 1996)
MixtureEM.tar.gz(archive)
MixtureEM.java
Mixture.java
Module.java
Database.java

Paper

• EM algorithm for more extended mixture model (not implemented by JAVA) can be obtained. See the paper ETL-TR-95-13E.pdf(79kbytes).
• Splus code for mixture modeling is available

• Mathematical analysis of a class of Gaussian mixtures is presented in the paper
  • ETL-TR-98-22.pdf.
  • The revised version is formally published in Neural Computation, Vol.12, No. 6, 1411-1428, 2000
  In short words, the contents of the paper showed "simple is not always best" mathematically.

-> Akaho's Home Page

Any questions should be made to
s . a k a h o @ a i s t . g o . j p