On Mon, Jan 18, 2010 at 4:42 PM, Ted Dunning <ted.dunn...@gmail.com> wrote:
> > DenseRandomizer is largely a performance optimization for the case where > the > number of probes equals the reduced dimensionality. It is probably of most > use for the fabled upcoming SVD decomposition code rather than for SGD. I guess Ted has figured out the trick to egging me into action! "Fabled" indeed, harumph! :) Initial patch to MAHOUT-180 has been posted. I'll clean up / disentangle the unit test (and make sure it runs with less than -Xmx1g), and get that committed so that people can play with it while I'm working on it (as it's now basically completely self contained, and doesn't touch much else). -jake
