[Topic-models] A problem on CTM program

cwarnold at ucla.edu cwarnold at ucla.edu
Mon May 5 13:10:40 EDT 2008


Hi Min,
I have this problem on an older Intel machine running Ubuntu Gutsy.  
Unfortunately I have not been able to solve it, but believe it does  
have to do with GSL. I have a newer Intel machine running Ubuntu Gutsy  
as well and have no problems running the CTM code on it. Both  
computers have identical GSL package installations. What is your  
setup? Does anyone know if this could somehow be hardware related?

Best,
-Corey


--------- Forwarded message ----------
From: "Min Qiu" <qium at cs.wisc.edu>
To: topic-models at lists.cs.princeton.edu
Date: Sun, 4 May 2008 22:04:43 -0500 (CDT)
Subject: [Topic-models] A problem on CTM program
Hi all,

These days I downloaded the CTM codes and tried to run it on my computer.
( I have redirected the link to local BLAS library). Unfortunately the
execution of program always halts on some iteration of EM. An example is
like this:

***** EM ITERATION 0 *****
doc     0   lhood -2.11049e+02   niter     5
doc     1   lhood -8.62415e+02   niter    20
doc     2   lhood -1.32725e+02   niter     5

I changed my data by removing some rows and run the CTM program
respectively, and found that sometime it went through the EM iteration 0,
but halted on iteration 1. Or it still halted on iteration 1, but can
print out different number of lines of doc.

I traced the execution of program and found it halts on a function vset()
which is implemented by calling the function gsl_vector_set() in GSl
library. I don't know what happend. Is there anybody who met the similar
problem before? Any suggestion would be appreciated!


best,
Min


More information about the Topic-models mailing list