boomsbloom / princeton-statistical-learning Goto Github PK

The Makefile in/dtm seems to be expecting there to be a file stl in 
/local/lib/, but no such file exists in dtm_release-0.8. I just removed the 
mention of /local/lib/stl in the Makefile and was able to compile successfully. 
I'm waiting for the sample DTM command to finish running, but it looks okay so 
far.

In the update_gamma function in lda.c file, it is overriding gamma using alpha 
in each iteration, everything is done in only one iter, is this correct? 

I also try to comment this line, and it is actually doing multiple iterations 
until convergence, but the likelihoods are keeping going down, which is not 
"optimizing" at all.

DTM compiles but segfaults immediately; the default of 0 for initialize_lda 
should instead be false

DTM does not output beta_t. I would like to get the topic proportions in each 
time slice. How can I get them?

Thanks