We don't have mega scale ols but we do have mega scale svd which
should be close to what you want if you have sparse data.
Sent from my iPhone
On Aug 20, 2010, at 1:37 PM, Chris Bates <[email protected]
> wrote:
Hi all,
I'm new to the list. I have a bunch of algorithms that I have
prototyped
during my PhD research in Matlab, but would like to apply them to some
datasets that are potentially larger than would fit into memory.
One very simple algorithm is a polynomial least square regression
that I'd
like to implement, but unsure where to start. In Matlab I've used 2
matrices, where one is transposed. I've also used a pseudo-inverse
function.
Can this algorithm be implemented in Mahout? Is it already
implemented? Are
there links that I could read that would point me towards that
answer? Are
there other packages I could use?? I've heard of UJMP and JAMA, but
haven't
used them. I also know python, so perhaps NumPy and SciPy would be
of help,
but doubt their scalability.
Thanks in advance for your help!
Chris
