Java Class for SVD of large sparse matrix (say 50,000 x 50,000) ?

Thread View: comp.lang.java.help

5 messages

[T] Back to Threads [V] Tree View [G] Group Articles

5 total messages Started by David Johnson Sat, 19 Jun 2004 12:59

#35

Author: David Johnson
Date: Sat, 19 Jun 2004 12:59

17 lines
1792 bytes

Hello,

What I am after is a GNU or similarly licensed class that will perform
Singular Value Decomposition on fairly large matrices (I'm experimenting
with Latent Semantic Analysis).

The Matrix Toolkit for Java (MTJ) http://www.math.uib.no/~bjornoh/mtj/ has a
SVD function, but that only seems to work up to about 1000 x 1000 elements
(with ~1GB RAM).  The Sparse Matrix Toolkit (SMT)
http://www.math.uib.no/~bjornoh/mtj/smt/ sounds promising, but hasn't got
SVD as a method.

You may have already guessed that linear algebra isn't my strong point!  Any
relevant pointers appreciated.

Thanks,
Dave.

Re: Java Class for SVD of large sparse matrix (say 50,000 x 50,000) ?

#36

Author: Jim Sculley
Date: Sat, 19 Jun 2004 19:51

25 lines
2309 bytes

David Johnson wrote:
> Hello,
> 
> What I am after is a GNU or similarly licensed class that will perform
> Singular Value Decomposition on fairly large matrices (I'm experimenting
> with Latent Semantic Analysis).
> 
> The Matrix Toolkit for Java (MTJ) http://www.math.uib.no/~bjornoh/mtj/ has a
> SVD function, but that only seems to work up to about 1000 x 1000 elements
> (with ~1GB RAM).  The Sparse Matrix Toolkit (SMT)
> http://www.math.uib.no/~bjornoh/mtj/smt/ sounds promising, but hasn't got
> SVD as a method.
> 
> You may have already guessed that linear algebra isn't my strong point!  Any
> relevant pointers appreciated.

JAMA has an SVD implementation.

http://math.nist.gov/javanumerics/jama/doc/

I cannot comment on its performance, only its existence.

Jim S.
-- 
Remove my extraneous mandibular appendages to reply via email.

Re: Java Class for SVD of large sparse matrix (say 50,000 x 50,000) ?

#37

Author: Roedy Green
Date: Sat, 19 Jun 2004 20:15

12 lines
1497 bytes

On Sat, 19 Jun 2004 15:51:54 -0400, Jim Sculley
<niceguyj@wisdomteeth.comcast.net> wrote or quoted :

>
>JAMA has an SVD implementation.

What sort of problems is SVD used for in the real world?

-- 
Canadian Mind Products, Roedy Green.
Coaching, problem solving, economical contract programming. 
See http://mindprod.com/jgloss/jgloss.html for The Java Glossary.

Re: Java Class for SVD of large sparse matrix (say 50,000 x 50,000) ?

#38

Author: Jim Sculley
Date: Sun, 20 Jun 2004 01:15

21 lines
2149 bytes

David Johnson wrote:
> Hello,
> 
> What I am after is a GNU or similarly licensed class that will perform
> Singular Value Decomposition on fairly large matrices (I'm experimenting
> with Latent Semantic Analysis).
> 
> The Matrix Toolkit for Java (MTJ) http://www.math.uib.no/~bjornoh/mtj/ has a
> SVD function, but that only seems to work up to about 1000 x 1000 elements
> (with ~1GB RAM).  The Sparse Matrix Toolkit (SMT)
> http://www.math.uib.no/~bjornoh/mtj/smt/ sounds promising, but hasn't got
> SVD as a method.


Actually, it does.

http://www.mi.uib.no/~bjornoh/mtj/doc/mt/fact/SingularvalueComputer.html

Jim S.
-- 
Remove my extraneous mandibular appendages to reply via email.

Re: Java Class for SVD of large sparse matrix (say 50,000 x 50,000) ?

#39

Author: David Johnson
Date: Sun, 20 Jun 2004 07:26

32 lines
2196 bytes

Jim,

I might be wrong, but isn't that part of  MTJ, rather than SMT?  I'll have
to look at it a bit more closely next week.

Dave.
"Jim Sculley" <niceguyj@wisdomteeth.comcast.net> wrote in message
news:qMGdnU5bSNiYeEndRVn-gw@comcast.com...
> David Johnson wrote:
> > Hello,
> >
> > What I am after is a GNU or similarly licensed class that will perform
> > Singular Value Decomposition on fairly large matrices (I'm experimenting
> > with Latent Semantic Analysis).
> >
> > The Matrix Toolkit for Java (MTJ) http://www.math.uib.no/~bjornoh/mtj/
has a
> > SVD function, but that only seems to work up to about 1000 x 1000
elements
> > (with ~1GB RAM).  The Sparse Matrix Toolkit (SMT)
> > http://www.math.uib.no/~bjornoh/mtj/smt/ sounds promising, but hasn't
got
> > SVD as a method.
>
>
> Actually, it does.
>
> http://www.mi.uib.no/~bjornoh/mtj/doc/mt/fact/SingularvalueComputer.html
>
> Jim S.
> -- 
> Remove my extraneous mandibular appendages to reply via email.

Thread Navigation

This is a paginated view of messages in the thread with full content displayed inline.

Messages are displayed in chronological order, with the original post highlighted in green.

Use pagination controls to navigate through all messages in large threads.

Back to All Threads