sylvester

Modern and expanded implementation of James Coglan's "Sylvester" matrix math library. The original project can be found at http://sylvester.jcoglan.com/

Documentation

The original documentation for "Sylvester" should help you through basic operations. An intro that contains node-specific features can also be found {on Chris Umbel's blog}[http://www.chrisumbel.com/article/sylvester_node_js_matrix_vector_math]. We're looking for someone to help get the documentation situation under control.

Installation

npm install sylvester

Usage

New Stuff

First I'd like to show some examples of features that aren't in the standard (non-node) Sylvester. I'll likely attempt to commit these back to Sylvester at some point soon.

Note that the decompositions are all available in pure JavaScript, but if the lapack NPM is installed with LAPACK built as a shared library then efficient native code will be used. The LAPACK integration is still highly experimental.

Vector

require('sylvester');
var a = $V([1, 2, 3]);

element-wise log:

console.log(a.log());

norm computation:

console.log(a.norm());

element-wise multiplication:

a.elementMultiply(vector);

element-wise division:

a.elementDivide(vector);

remove first n nodes:

a.chomp(n);

return vector with first n nodes:

a.top(n);

add all elements into a single scalar:

a.sum()

multiply all elements into a single scalar:

a.product()

return a vector with the elements parameter on the bottom:

a.augment(elements)

Matrix

var A = $M([[1, 2, 3], [4, 5, 6]]);

return subset of rows, columns:

// startRow, endRow, startCol, endCol
A.slice(2, 3, 2, 3);

divide matricies:

A.div($M([[0.5, 1], [1, 2], [2, 3]]));

scalar addition/subtraction

A.add(1);
A.subtract(1);

element-wise log:

console.log(A.log());

element-wise multiplication:

A.elementMultiply(vector)

add all elements into a single scalar:

A.sum()

returns a vector of the indexes of maximum values ([3 3]):

$M([[1, 2, 3], [5, 4, 6]]).maxColumnIndexes()

returns a vector of minimum column indexes ([1 2]):

$M([[1, 2, 3], [5, 4, 6]]).minColumnIndexes();

returns a vector of max values ([3 6]):

$M([[1, 2, 3], [5, 4, 6]]).maxColumns()

returns a vector of minimum values ([1 4]):

$M([[1, 2, 3], [5, 4, 6]]).minColumns()

create a 2x3 matrix of ones:

var Ones = Matrix.One(2, 3);

LU decomposition (with partial pivoting)

var lu = A.lu(); console.log(lu.L); console.log(lu.U); console.log(lu.P);

QR decomposition (feature still inefficient and experimental, but uses pure javascript):

var qr = A.qr();
console.log(qr.Q);
console.log(qr.R);

SVD decomposition (feature still inefficient and experimental, but uses pure javascript):

var svd = A.svd();
console.log(svd.U);
console.log(svd.S);
console.log(svd.V);

PCA

var A = $M([[1, 2], [5, 7]]).pcaProject(1).eql($M([
            [-2.2120098720461616],
            [-8.601913944732665]
        ]);
var pca = A.pcaProject(1);
var Z = pca.Z;
var A = Z.pcaRecover(pca.U);

Solving systems of equations

// sovle Ax = b for x
var A = $M([[2, 4], [2, 1]]);
var b = $V([1, 0]);
console.log(A.solve(b));

== Old Stuff

Below is a basic illustration of standard matrix/vector math using the standard Sylvester API. This documentation is rather incomplete and for further details please consult {the official sylvester API documentation}[http://sylvester.jcoglan.com/docs] at http://sylvester.jcoglan.com/docs.

Vectors

require('sylvester');

create two vectors:

var a = $V([1, 2, 3]);
var b = $V([2, 3, 4]);

compute the dot product:

var r = a.dot(b);

add two vectors:

var c = a.add(b);

multiply by scalar:

var d = a.x(2);

Matrices

require('sylvester');

create two matrices:

var A = $M([[1, 2], [3, 4]]);
var B = $M([[1, 2, 3], [4, 5, 6]]);

multiply the matrices:

var C = A.x(B);

transpose a matrix:

var B_T = B.transpose();
// B is 2x3, B_T is 3x2

Contributing

Contributions to node-sylvester are more than welcome. If you changed or added functionality, you should make sure tests pass. You can do this locally by running npm test, after running npm install initially.

Development depends on the lapack module, which requires the LAPACK linear algebra library to be installed locally.

• OSX seems to have lapack installed by default.

• On Linux, lapack is available via most package managers, including apt.

• On Windows, the easiest way is to use Docker. Build the Dockerfile in this repository once: docker build -t node-sylvester .. Then:

  • On the default shell: docker run --rm -v \pwd`:/sylv -w /sylv node-sylvester sh -c /bin/watch`
  • On Cygwin: docker run --rm -v \cygpath -aw $(pwd)`:/sylv -w /sylv node-sylvester sh -c /bin/watch`