UJMP» Universal Java Matrix Package »

UJMP v0.2.4 released

admin — Mon, 08 Feb 2010 11:17:53 +0000

It’s been a long time since the last release. This version brings a lot of improvements, especially concerning linear algebra functions. UJMP now fully supports solving linear equations and decomposition methods: Singular value decomposition, Eigenvalue decomposition, LU, QR, and Cholesky decomposition.
Many basic functions such as scaling and matrix multiplication can use multiple threads which improved performance a lot.

UJMP v0.2.3 released

admin — Sat, 04 Jul 2009 18:53:38 +0000

This version brings support for BigInteger and BigDecimal matrices. Also a lot of refactoring, bug fixing and other improvements.

UJMP v0.2.2 released

admin — Sun, 08 Mar 2009 16:50:38 +0000

This is a maintenance release which includes changes for the next version of JDMP.
New Factory classes have been created to allow easier switching of matrix implementations in the future.
A cool demo showing the Mandelbrot fractal has been included
Some minor bugfixes, improvements and API changes.

UJMP v0.2.1 released

admin — Sat, 21 Feb 2009 23:46:17 +0000

This release brings support for additional matrix packages and other libraries:

Interface to Shared Scientific Toolbox
Interface to Ojalgo
Interface to parallelcolt
Interface to JSci
Interface to JScience
Interface to Lucene

A benchmark is included to compare the performance of UJMP against other matrix libraries, including Colt, Jama, and many others.

Additional interfaces have been created to better reflect the the matrix type, e.g. SparseMatrix.

And of course: bug fixes and improvements!

Version 0.2.0 (2009-01-18)

admin — Sun, 18 Jan 2009 23:57:46 +0000

dependencies removed: ujmp-core and ujmp-gui require no other jar files!
refactoring of interfaces: cleaner class hierarchy
included new methods to replace missing values
Main methods for core and GUI to display information
a lot of bugfixes

Display a Matrix on the Screen

admin — Sun, 09 Nov 2008 22:17:00 +0000

Every matrix can be visualized in a JFrame by invoking:

matrix.showGUI();

Import a Huge CSV Matrix

admin — Sun, 09 Nov 2008 21:38:33 +0000

If you have a very large CSV file, that is bigger than the available memory, you can link to the file instead of importing the data:

Matrix hugeMatrix = MatrixFactory.linkToFile(FileFormat.CSV, new File("VeryBigFile.csv"));

However, this matrix will be read-only.

Collections

admin — Thu, 09 Oct 2008 12:42:55 +0000

You can also create a ListMatrix instead of an ArrayList. A ListMatrix behaves like a java.List but is a Matrix at the same time:

ListMatrix S = new DefaultListMatrix();
// List methods:
S.add(0.5);
S.clear();
// Matrix methods:
S.exportToFile(someFile);
S.showGUI();
Matrix P = S.plus(5).times(2);
// but: P is not a ListMatrix anymore

The same works also for maps:

MapMatrix map = new DefaultMapMatrix(); map.put(key,o);

I need a certain Matlab function, where do I find it in UJMP?

admin — Thu, 09 Oct 2008 12:32:43 +0000

UJMP supports many Matlab functions, but not all.
But you can execute Matlab from within UJMP when you use Linux:

Matlab.getInstance().setMatrix("matrixInMatlab", ujmpMatrix);
Matlab.getInstance().execute("result = matlabFunction(matrixInMatlab)");
Matrix m = Matlab.getInstance().getMatrix("result");

Unfortunately, this is not working in Windows.

You might have to tell UJMP through a parameter, where the program is installed:
-DMatlab="/opt/matlab/matlab"

Can I resize a matrix?

admin — Thu, 09 Oct 2008 12:27:11 +0000

Resizing depends on the matrix implementation. A dense matrix is fixed in size, as the data is stored in a double[][] array.

However, some matrix implementations allow resizing. For example, you can create a sparse matrix with MatrixFactory.sparse(0, 0) and then change the size dynamically with matrix.setSize(5, 5). Note that sparse matrices are much slower than dense matrices.

It might be better to append one matrix to another for resizing:

Matrix matrix = Matrix.zeros(5, 5);
Matrix newRow = Matrix.zeros(1, 5);

Matrix biggerMatrix = matrix.appendVertically(newRow);

Matrix biggerMatrix = MatrixFactory.vertCat(matrix, newRow);