Matrix implementations

Overview

core.matrix supports multiple implemntations of vector/matrix maths conforming to the same API.

Implementations included in core.matrix

• persistent-vector : regular Clojure vectors work as core.matrix implementations
• double-array : Java double[] arrays work as 1D mutable vectors
• ndarray : a general propose mutable n-dimensional array, stores objects of any type
• sequence : Clojure sequences (anything implementing ISeq) can be used

Main external core.matrix implementations:

• vectorz-clj : An implementation based on the Vectorz library, which offers fast pure-JVM matrix and vector computation with double-valued matrices. Designed for games, simulations and machine learning. Particularly efficient for small matrix and vector maths.
• Clatrix : A wrapper of jblas which in uses the BLAS native libraries for matrix computation. Offers high performance for large matrices, at the cost of needing a native code dependency.

Work in progress / experimental implementations

• Parallel Colt - Support for Parallel Colt matrices
• Cerebro - inital support for EJML
• image-matrix - Support for treating a Java BufferedImage as a matrix of pixel colour data.

Example implementations we might want to support in the future:

• Apache Commons maths
• javax.vecmath (part of Java3D)
• UJMP