csr-matrix

Compressed sparse row matrix class

csr-matrix

A very minimal incremental compressed sparse row matrix library for JavaScript.

Usage

First install using npm:

npm install csr-matrix

Then you can create a matrix and apply it to a vector like this:

var dict = {}
dict[[1,2]] = 1
dict[[0,0]] = 2.5
dict[[3,4]] = 5
 
var CSRMatrix = require("csr-matrix")
 
var M = CSRMatrix.fromDictionary(dict)
console.log(M.apply([1,2,3,4,5]))

CSRMatrices are optimized for exactly one thing: fast matrix-vector multiplies. The way you do this is by calling the following method:

This computes the normal matrix-vector product, but is often much faster than a dense multiply since the matrix is stored in a compressed sparse format.

  • vector is the vector to be multiplied
  • result is an optional array which gets the result. If not specified, a new vector gets allocated

Returns the resulting product

There are several ways to create csr-matrices. The most direct way to do this is to just call the constructor yourself:

Where:

  • rows is an array of row indices
  • row_ptrs is an array of pointers to the start of each row
  • columns is an array of column names
  • column_data is a pointer to the start of each column's run
  • data is an array of all the entries of the matrix stored left-to-right and top-to-bottom

Calling this method directly is not advised. Instead, you should use one of the more user-friendly constructors:

Turns an array of entries of the form [row, column, value] into a sparse matrix. Note that if there are some zero rows at the end of the matrix, you need to specify the number of rows in the nrows optional argument.

Converts a JavaScript object with entries for the form "row,column" into a sparse matrix.

Turns an array-of-arrays into a csr matrix

Turns a 2D dimensional ndarray into a csr matrix

Credits

(c) 2013 Mikola Lysenko. BSD