checkout the source code from:
svn checkout http://svn.apache.org/repos/asf/lucene/mahout/trunk
Lucene/mahout/trunk
Chaitanya
On Feb 2, 2008 10:01 PM, Jason Rennie <[EMAIL PROTECTED]> wrote:
> How do you get the code? SVN is asking for username/password...
>
> What is the constant hit you take
How do you get the code? SVN is asking for username/password...
What is the constant hit you take for using HBase vs. an in-memory sparse
matrix library?
Jason
On Thu, Jan 31, 2008 at 2:39 AM, Lukas Vlcek <[EMAIL PROTECTED]> wrote:
> BTW: Have you seen http://wiki.apache.org/hadoop/Matrix yet?
This is a good idea. Many of these things will be needed in Hadoop
for others, so ideally, we can try to help along some of these things
in Hadoop by testing them out, patching, etc. This will help keep our
deps. down and make Hadoop better, as well.
-Grant
On Jan 31, 2008, at 2:39 AM, L
BTW: Have you seen http://wiki.apache.org/hadoop/Matrix yet?
On Jan 30, 2008 11:23 PM, Jason Rennie <[EMAIL PROTECTED]> wrote:
> On Jan 30, 2008 2:49 PM, Ted Dunning <[EMAIL PROTECTED]> wrote:
>
> > For efficient row and column access, I built a hybrid that has both CSR
> > and
> > CSC represenat
On Jan 30, 2008 2:49 PM, Ted Dunning <[EMAIL PROTECTED]> wrote:
> For efficient row and column access, I built a hybrid that has both CSR
> and
> CSC represenations under the hood. I also don't care much about mutation.
Cool :) OS? Or, something you built for VEOH? Probably best to move
off-
For efficient row and column access, I built a hybrid that has both CSR and
CSC represenations under the hood. I also don't care much about mutation.
Relative to API, I think we need to support (at least):
A * v
v * A
A * A
A' * A
rowSum(A)
columnSum(A)
sum(A)
forEachNo
On Jan 30, 2008 1:58 PM, Ted Dunning <[EMAIL PROTECTED]> wrote:
> Many of my matrices need both fast column AND fast row access. They are
> also binary. Yet another implementation.
>
Do you know a good storage format where both column and row access is
efficient? Hadn't found a good one yet, b
On 1/30/08 10:37 AM, "Jason Rennie" <[EMAIL PROTECTED]> wrote:
> Btw, not familiar w/ RSR. Didn't mean CSC, did you?
I did mean CSC. Forgot that the last letter was row/column choice.
Many of my matrices need both fast column AND fast row access. They are
also binary. Yet another impleme
On Jan 30, 2008 12:40 PM, Ted Dunning <[EMAIL PROTECTED]> wrote:
> I wrote my own package as well based roughly on Colt's high level
> structure.
Has a sparse matrix code-based been chosen for Mahout? If not, we might as
well start the discussion :)
It turns out that for many data-mining proje
I wrote my own package as well based roughly on Colt's high level structure.
It turns out that for many data-mining projects you need several flavors of
sparse matrices ([CSR, RSR] x [binary values, real values, complex values] x
...).
It also helps a lot of have row and column labels a la R.
10 matches
Mail list logo
