Java matrix libraries [closed]

Question

I was wondering whether any of the well-known matrix libraries for Java, such as Colt or EJML, actually provide similar functionality as MatLab? For instance, I can't seem to find anywhere in the definition of their API simple method to add or subtract two matrices/vectors by each other, which seems to be the most common operation used. Am I missing something?

Ovolo

Answer 1

Try Apache Commons Math library. org.apache.commons.math3.linear package contains the functions that you want. Home page

Falsetto

Yes, but that is probably less efficient then Colt or EJML – Ovolo

Answer 2

Some Java libraries for linear algebra are:

• Apache Commons Math: http://commons.apache.org/proper/commons-math/
• jeigen - a wrapper for eigen - https://github.com/hughperkins/jeigen (includes complex and rarely found feature like matrix exponential and matrix logarithm)
• jblas http://mikiobraun.github.io/jblas/ (also features more complex functions like matrix exponential, also very fast).
• Colt http://acs.lbl.gov/software/colt/
• JAMA http://math.nist.gov/javanumerics/jama/
• UJMP - http://sourceforge.net/projects/ujmp/

EDIT maybe we can extend this list whenever one comes across and you know - the world keeps moving:

• ojAlgo - http://ojalgo.org/ has promising benchmarks
• Efficient Java Matrix Library (EJML) - http://ejml.org
• ParallelColt - https://sites.google.com/site/piotrwendykier/software/parallelcolt
• la4j - http://la4j.org/
• MTJ - https://github.com/fommil/matrix-toolkits-java
• nd4j - https://nd4j.org/ lets you choose underlying native implementations like cuda or openBlas

Note: Personally: I use Apache Commons Math and Colt in my own project (http://www.finmath.net). While commons math is actively developed, I found that Colt is still faster in some tasks (like Eigenvalue decomposition). For that reason I use some kind of wrapper which allows me to quickly switch the underlying library (I only need a few things like solving systems of equations and Eigenvalue decomposition).

Spook

JEigen does not appear to implement solving sparse matrices so I did it myself. – Zincography

jscience.org/api/org/jscience/mathematics/vector/… -- JScience also supports basic math operations on Matrices and Vectors – Affright

github.com/BIDData/BIDMat might be a good addition as well.. (though I don't have experience with it) – Loxodromic

Answer 3

The interface for COLT gives you a generic OP: assign(matrix, function), which you can use to add or subtract matrices and vectors.

As the javadocs for assign() says:

Assigns the result of a function to each cell; x[row,col] =function(x[row,col],y[row,col]).

So by using using an addition function as function - you can add matrices.

Tacita

Answer 4

There is also la4j library that supports functional programming features as functors. You can use transform() method with manual hi-order function.

Conic

Answer 5

You can use this one. It is very simple and supply basic matrix operations.

You can add/substract/multiply etc 2d matrixes and vectors.

Gurdwara

Answer 6

You may try my newly Java library (compatible with Java 6+) that can be used for all sorts of applications notably for performing multi-threaded matrix calculations (extending JAMA with multi-threading; currently 30% faster than the available open source libraries). You can check it out on GitHub:

https://github.com/b-io/io.barras/tree/master/java/jupiter

Note that each project is independent from each other and is under the very permissive MIT License.

Etch