Automatic

[owenchang]

Hello,

The current parameter learning function "Data -> Learning Parameters" in Genie requires MANUAL mapping of network variables (and states) and data variables (states).

Is there some way to easily automate this process in Genie or can only be done using SMILE APIs (then I don't see a parameter learning routine I can just call there..)

Thanks a millions,

Owen

[shooltz[BayesFusion]]

The current parameter learning function "Data -> Learning Parameters" in Genie requires MANUAL mapping of network variables (and states) and data variables (states).

GeNIe tries to match as many nodes/variables as possible by comparing the identifiers. The states are also matched. This will of course work only if the identifiers of nodes/outcomes and variables/values are equal.

Is there some way to easily automate this process in Genie or can only be done using SMILE APIs (then I don't see a parameter learning routine I can just call there..)

There's no way to add more automation (like more sophisticated matching in GeNIe). You may consider changing the network and/or data file to match - this will allow auto-matching in GeNIe to perform all the work.

The learning code is distributed in SMILearn:
http://genie.sis.pitt.edu/wiki/Reference_Manual:_DSL_em