Create Sparse Matrix with Coordinate Storage System?

Question

I am writing a java program which involves working with a 1058 X 1058 matrix containing float values. This matrix contains many zero values and so I need to store this as a sparse matrix and later use this matrix to generate a spanning tree. I checked many popular libraries like Colt, Jama, but somehow I am unable to put them to work with my code. I would like to have a coordinate storage system (similar to obtained in matlab using the sparse() function) like this:

(1055,1045)    1.0000
(1056,1045)    1.0000
(1057,1045)    1.0000
(1058,1045)    1.0000
(1047,1046)    1.0000
(1048,1046)    1.0000
(1049,1046)    1.0000
(1050,1046)    1.0000
(1051,1046)    1.0000
(1052,1046)    1.0000
(1053,1046)    1.0000
(1054,1046)    1.0000
(1055,1046)    1.0000

Can anyone suggest how to go about this?

_" I checked many popular libraries like Colt, Jama, but somehow I am unable to put them to work with my code."_ If you give more detail about the problems you had with those libraries, we could help you get them working. — Kevin, Jul 09 '13 at 19:43
I am trying to use the methods of the class SparseDoubleMatrix2D (http://acs.lbl.gov/software/colt/api/cern/colt/matrix/impl/SparseDoubleMatrix2D.html#assign(cern.colt.matrix.DoubleMatrix2D). But I am not able to generate the coordinate storage wise output using the methods listed there. — novicegeek, Jul 09 '13 at 20:26

David · Accepted Answer · 2013-07-09T21:19:13.367

0

You can do it quickly without any lib. Create the following class :

MatrixIndex implements Comparable<MatrixIndex>
{
  private final int _x;
  private final int _y;

  ...
}

Then use it in some :

TreeMap<MatrixIndex,Double>

Cheers

edited Jul 09 '13 at 21:19

answered Jul 09 '13 at 21:00

David

1,138
5
15

Thank you for your reply. I read more on comparator and TreeMap and have got an idea of how to implement it. But, as for the class MatrixIndex, is a comparator required in this case? As far as I understand, TreeMap gives returns data in sorted manner and since my return value will be coordinates and value, which are numbers so they will be already sorted. If I use a comparator as you suggested, how should I get the float[][] matrix values into this? Thanks in advance – novicegeek Jul 11 '13 at 17:17
I was able to get the desired output using TreeMap. Thank you for your help! – novicegeek Jul 11 '13 at 18:45

Vladimir Kostyukov · Answer 2 · 2013-07-15T11:15:36.863

There is a la4j (Linear Algebra for Java) library that handles this with CRSMatrix/CCSMatrix sparse matrix types and MatrixMarketStream class. Here is the brief example:

Source file "matrix.mm" (5x5 matrix with 8 non-zero elements):

%%MatrixMarket matrix coordinate real general
5  5  8
1     1   1.000e+00
2     2   1.050e+01
3     3   1.500e-02
1     4   6.000e+00
4     2   2.505e+02
4     4  -2.800e+02
4     5   3.332e+01
5     5   1.200e+01

Java source (la4j usage):

Matrix a = new CRSMatrix(Matrices.asMatrixMarketSource(
                           new FileInputStream(new File("matrix.mm"))));

System.out.println("DET(A) = " + a.determinant());

About the MatrixMarket format read here. About CRS/CCS formats read here and here.

To generate MatrixMarket output use following code:

Matrix a = new CRSMatrix(...);
MatrixStream out = new MatrixMarketStream(new FileOutputStream(
                        new File("matrix.mm")));

out.writeMatrix(a);

Create Sparse Matrix with Coordinate Storage System?

2 Answers2