Convergence criteria for EM algorithm?

The front end.
Post Reply
nisha269
Posts: 5
Joined: Mon Jun 27, 2011 5:31 pm

Convergence criteria for EM algorithm?

Post by nisha269 »

Hi,
I would like to know as to what is the convergence criteria that is being used in GeNIe to learn the parameters using an EM algorithm.

My qns.
1. Is it just fixed to a large no. so the algorithm runs for so many no. of iterations or some other criteria? How do i change this criteria if needed??
2. What is the significance of the Log p value that is returned after learning the network parameters??

Thanks.
mark
Posts: 179
Joined: Tue Nov 27, 2007 4:02 pm

Re: Convergence criteria for EM algorithm?

Post by mark »

1. It looks at the relative improvement of the log likelihood. When it's below a certain threshold, learning is stopped. This criteria cannot be changed at the moment.

2. It is the data log likelihood of the learned parameters.
Post Reply