[Topic-models] Multi threaded extension of David Blei's LDA

M. Edward (Ed) Borasky znmeb at borasky-research.net
Tue Jan 4 00:25:00 EST 2011


On Mon, 3 Jan 2011 21:01:05 -0800, Ramesh Nallapati
<nmramesh at cs.stanford.edu> wrote:
> Hi all,
> 
> I had built a multi-threaded extension of Prof. David Blei's LDA-C code
> more
> than 3 years ago, and I recently used it on a corpus of half-a-million
> documents with 150 topics on an 8 core machine with 24G memory, and it
> finished in less than a day's time. I remember a discussion thread about
> scalable implementations of LDA a while ago in this list, and I thought
> people on this list might find this implementation useful. You may
download
> the code from https://sites.google.com/site/rameshnallapati/software.
The
> tar bundle contains a README file that contains usage instructions, that
> are
> pretty much like Prof. Blei's LDA code except for an extra parameter
that
> lets you specify the number of threads to run.
> 
> I welcome any feedback, comments and suggestions on the implementation.

Looks good to me - do you have a small test dataset I can use to verify
that it's functioning correctly?
> 
> Thanks
> -Ramesh

-- 
http://twitter.com/znmeb http://borasky-research.net

"A mathematician is a device for turning coffee into theorems." -- Paul
Erdős


More information about the Topic-models mailing list