Fwd: RE: CSparse package

[boisvert@nist.gov]

Sender: "Cleve Moler" <Cleve.Moler@mathworks.com>
Subject: RE: CSparse  package


Hi --

Tim Davis develops software for solving sparse systems of linear
equations.  He does not work on eigenvalue problems.  The MATLAB sparse
eigenvalue functions "eigs" and "svds" are based on Danny Sorensen's
Arnoldi package, www.caam.rice.edu/software/ARPACK/.  I do not know of
any packages in Java for sparse matrix computations of any kind, linear
equations or eigenvalue problems.

  -- Cleve Moler 

-----Original Message-----
From: jama@nist.gov [mailto:jama@nist.gov] On Behalf Of Sione
Sent: Thursday, August 03, 2006 2:12 PM
To: Multiple recipients of list
Subject: CSparse package

Hello Tim Davis,

I have downloaded your  C/Matlab package for sparse matrix computation 
found here to play with:

http://www.mathworks.com/matlabcentral/fileexchange/loadFile.do?
objectId=11740&objectType=FILE

I will order a copy of your upcoming book ("/Direct Methods for Sparse 
Linear Systems/") when it becomes available, but just a quick question 
regarding your package. 

What are the functions or equivalent functions  from your package that 
correspond to  Matlab   'eigs'   and   'svds'  ?  I am no expert in  C

but  I intend to port your  sparse package to Java and extend  JAMA 
capability with sparse classes as SparseEigenDecomposition ,  
SparseSingularValueDecomposition, etc. I find that analysis as  market 
basket in data-mining  cannot be done via  dense-matrix  JAMA  as huge 
"customer-by-item"  transaction matrix using non-negative matrix 
factorisation (NNMF) will have high computational cost.   
"Customer-by-Item"  transaction matrix is sparse therefore I think that 
it should be done via sparse computation.

Any hint would be appreciated.

Thanks,
Sione.

//




----- End forwarded message -----






----- End forwarded message -----