vectorious
TypeScript icon, indicating that this package has built-in type declarations

6.1.12 • Public • Published

Vectorious Logo

A linear algebra library, written in TypeScript and accelerated with C++ bindings to BLAS and LAPACK.

Installation

Follow the installation instructions in nlapack and nblas to get maximum performance.

# with C++ bindings
$ npm install vectorious

# or, if you don't want C++ bindings
$ npm install vectorious --no-optional

There are three output bundles exposed in this package.

CommonJS

A node.js bundle, can be found in dist/index.js and imported with the require() syntax:

const v = require('vectorious');

Browser

A browser bundle, can be found in dist/index.browser.js and imported with the <script> tag:

<script src="dist/index.browser.js" />

It exposes a global variable named v in the window object and can be accessed like this:

<script>
  const x = v.array([1, 2, 3]);
</script>

ES module

Added in version 6.1.0, vectorious exposes an ES module bundle at dist/index.mjs which can be imported using the import syntax:

import { array } from 'vectorious';

const x = array([1, 2, 3]);

Usage

Unless stated otherwise, all operations are in-place, meaning that the result of the operation overwrites data in the current (or in the static case leftmost) array. To avoid this, an explicit copy call is needed before the operation (copy(x) or x.copy()).

import { array, random, range } from 'vectorious';

// Create a random 2x2 matrix
const x = random(2, 2);
/*
array([
  [
    0.26472008228302,
    0.4102575480937958
  ],
  [
    0.4068726599216461,
    0.4589384198188782
  ]
], dtype=float64)
*/

// Create a one-dimensional vector with values from
// 0 through 8 and reshape it into a 3x3 matrix
const y = range(0, 9).reshape(3, 3);
/*
array([
  [ 0, 1, 2 ],
  [ 3, 4, 5 ],
  [ 6, 7, 8 ]
], dtype=float64)
*/

// Add the second row of x to the first row of x
y.slice(0, 1).add(y.slice(1, 2));
/*
array([
  [ 3, 5, 7 ],
  [ 3, 4, 5 ],
  [ 6, 7, 8 ]
], dtype=float64)
*/

// Swap the first and second rows of x
y.swap(0, 1);
/*
array([
  [ 3, 4, 5 ],
  [ 3, 5, 7 ],
  [ 6, 7, 8 ]
], dtype=float64)
*/

// Create a 2x2x1 tensor
const z = array([
  [[1], [2]],
  [[3], [4]],
]);
/*
array([
  [ [ 1 ], [ 2 ] ],
  [ [ 3 ], [ 4 ] ]
], dtype=float64)
*/

Documentation

Examples

Basic

Machine learning

Testing

All functions are accompanied with a .spec.ts file.

The Jest testing framework is used for testing and the whole test suite can be run using a single command:

$ npm test

Benchmarks

All functions are accompanied with a .bench.ts file.

Run all benchmarks with:

$ npm run benchmark

Or for a single function with:

$ npx ts-node src/core/abs.bench.ts

Dependents (10)

Package Sidebar

Install

npm i vectorious

Weekly Downloads

81

Version

6.1.12

License

MIT

Unpacked Size

804 kB

Total Files

104

Last publish

Collaborators

  • waylonflinn
  • mateogianolio