EM Learn

[oscarpc]

Thank you so much.
If I understood correctly then, I have to change the DBN structure for the time being so the program can work. In particular, I have to remove the initial conditions. Am I right?

This is what I did. I created a new Network with no anchor nodes (see attachment). Nonetheless, the SMILE/SMILearn program gets stuck in the very same line as before and no EM learning occurs.
However, the very same new Network seems to work in GENIE.

My interpretation is that there is an error in addition to the aforementioned with the anchors nodes. I hope this is an error from my part. Again, I would be very grateful if you could help. Please, see attachments.

Thank you very much for your help. I really appreciate it.
Kind regards,

Oscar PC

[shooltz[BayesFusion]]

GeNIe and SMILE build for OSX have been refreshed. EM has relaxed constraints on the structure of the DBN and incoming anchor (initial condition) arcs are supported as long as there's incoming temporal arc. You should be able to use EM with DBN_v07.xdsl and data31.txt.

I didn't check your network/data matching code yet. You may consider using DSL_idArray::FindPosition to simplify it a bit. In any case, add some output when matching so you can see what's going on.

[oscarpc]

Thank you very much!
I will try the new version.
When you say "GeNIe and SMILE build for OSX have been refreshed", is there a version for GENIE for OSX?
Kind regards,
Oscar

[shooltz[BayesFusion]]

When you say "GeNIe and SMILE build for OSX have been refreshed", is there a version for GENIE for OSX?

No, GeNIe runs natively on Windows only. The OSX part applies only to SMILE.