As a general comment -- I can't imagine that anything beats an optimized, special purpose solution for a problem. It's not going to be possible to replicate that on a general purpose tool like Hadoop in volunteer time. But I bet a few tricks and optimizations, perhaps gleaned from this, could make a lot of difference.
On Wed, Jun 29, 2011 at 4:20 PM, Ian Upright <[email protected]> wrote: > I also wonder what memory limitations it may have as compared to the Mahout > implementation. (with regards to number of terms/topics/documents) > > Ian
