• Status: Solved
• Priority: Medium
• Security: Public
• Views: 1676

# Expectation Maximization algorithm for gamma mixture model

I'm looking for an Expectation Maximization(EM) algorithm for estimating paramers of gamma mixture model . The distribution is givin in the image
0
00matrix00
• 4
• 4
1 Solution

Commented:
There's only one Expectation Maximization algorithm. It's explained fairly well here http://www.cc.gatech.edu/~dellaert/em-paper.pdf.
0

Author Commented:
Actually what I want is how I update the parameters required for the gamma distribution in the Maximization step of EM
0

Commented:
how I update the parameters
You just start with an initial guess for the expectation step and the maximization step gives the new guess. So you just plug the new guess in.
If that's not what you mean, then can you give more detail including you expectation step?
0

Author Commented:
I mean that I want the exact formula for updating these two parameters
0

Commented:
Can you post what you have so far (especially your Expectation step) so that I don't go a different direction and cause confusion?
0

Author Commented:
In the Expectation step, I should compute the member ship probability using the form (Attached ). How can I update the three paramters of gamma mixture model in the maximization step.

Thanks
0

Author Commented:
Thank U TommySzalapski,
Actually, I found the formula for updating the parameters. The problem now is how to generate an initial value for these paramters. I trying to use random intialization for them but when I implemented , the pdf generated using EM didn't the fitted data histogram.
0

Commented:
Random initialization is usually not good. Just pick an initial guess of .5 or something. You know your application, so just pick a guess that you think makes sense. Small positive numbers usually are safe bets to use as guesses. I would avoid guessing 0 for most things just in case.
0
Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.