> I don't see anything wrong offhand.  You might look at MAHOUT-399.  I think 
> we are trying to review how LDA performs at the moment.  From what I 
> understand, you aren't guaranteed the same results each time (I wonder if 
> there is a way to at least provide some sort of seed value so that one can 
> reproduce a set of results)
>
> At any rate, it's good that you put up detailed instructions of what you did, 
> so that we can compare them.

thanks. i might start on a smaller set with clear distinct topics to
make sure my steps are sane and then build up. i'll let you know how i
go.

Reply via email to