[Topic-models] [Mallet][hPAM]To calculate the empirical loglikelihood for held-out data in hPAM model

Peixian Chen sbbisabel at gmail.com
Thu Nov 10 09:46:21 EST 2016


I am now using the Mallet Java api to build hierarchical PAM model. There
is a function
*public void estimate (InstanceList documents, InstanceList testing,int
numIterations, int showTopicsInterval,int outputModelInterval, int
optimizeInterval, String outputModelFilename,Randoms r),*

which seems to build the model and to estimate the test data also. However,
the "testing" part is not really used. Would anybody please kindly inform
me of the way to calculate Empirical Loglikelihood for held-out data using
the hPAM model? Thanks a lot.

Best Regards,
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cs.princeton.edu/pipermail/topic-models/attachments/20161110/2cc08629/attachment.html>

More information about the Topic-models mailing list