Hello, dear
How can i estimate mutual information or entropy value for the network?
Thanks in advance
TAN model
-
marek [BayesFusion]
- Site Admin
- Posts: 455
- Joined: Tue Dec 11, 2007 4:24 pm
Re: TAN model
Hi Islam,
Mutual information can be calculated for a pair of variables and this can be calculated in GeNIe using the programs diagnostic extensions. Please have a look at the manual chapter of support for diagnosis. By "entropy value for the network" you means entropy of the underlying joint probability distribution? GeNIe does not have such a calculation built in but it would also be challenging. Please note that a joint probability distribution of n binary variables contains 2^n elements. In calculating entropy one would in general have to look at all these elements. You can probably do it for a small network (10-20 variables) but not in general.
I hope this helps,
Marek
Mutual information can be calculated for a pair of variables and this can be calculated in GeNIe using the programs diagnostic extensions. Please have a look at the manual chapter of support for diagnosis. By "entropy value for the network" you means entropy of the underlying joint probability distribution? GeNIe does not have such a calculation built in but it would also be challenging. Please note that a joint probability distribution of n binary variables contains 2^n elements. In calculating entropy one would in general have to look at all these elements. You can probably do it for a small network (10-20 variables) but not in general.
I hope this helps,
Marek