On Mon, Dec 5, 2011 at 4:17 PM, Chris Grier <[email protected]> wrote: > > I can certainly try out the new implementation. Now that the example works > I'll be working with other data. If you let me know how to use the new > implementation, I'll go ahead and start in with that. >
Yep, if you start with "./bin/mahout cvb0 -h" you'll get the help output for the "Collapsed Variational Bayes (0th derivative approximation)" [ie cvb0], which is the new LDA impl. Its got most of the same input args as the old one, but a few new args, if you can try it out, and report back on what seems confusing or not sensible, I'll know where to focus my documentation! Either way, I'll try to copy-n-paste a (corporate internal) confluence page onto Mahout's wiki, which details the way we've been using it internally. -jake
