Distributions-normal-random NPM

Normal Random Variables

Creates a matrix or array filled with draws from a normal distribution.

Installation

$ npm install distributions-normal-random

For use in the browser, use browserify.

Usage

var random = require( 'distributions-normal-random' );

random( dims )

Creates a matrix or array filled with draws from a normal distribution. The dims argument may either be a positive integer specifying a length or an array of positive integers specifying dimensions. If no dims argument is supplied,the function returns a single random draw from a normal distribution.

var out;

// Set seed
random.seed = 2;

out = random( 5 );
// returns [ ~-0.832, ~0.735, ~-1.432, ~0.057, -0.13 ]

out = random( [2,1,2] );
// returns [ [ [~0.969,~-0.394] ], [ [~0.599,~-1.511] ] ]

The function accepts the following options:

mu: mean parameter. Default: 0.
sigma: standard deviation. Default: 1.
seed: positive integer used as a seed to initialize the generator. If not supplied, uniformly distributed random numbers are generated via an underlying generator seedable by setting the seed property of the exported function.
dtype: output data type (see matrix for a list of acceptable data types). Default: generic.

The normal distribution is a function of two parameters: mu(mean) and sigma > 0(standard deviation). By default, mu is equal to 0 and sigma is equal to 1. To adjust either parameter, set the corresponding option.

var out = random( 5, {
	'mu': 20,
	'sigma': 4,
});
// returns [ ~25.293, ~21.105, ~21.347, 22.57, ~18.726 ]

To be able to reproduce the generated random numbers, set the seed option to a positive integer.

var out = random( 3, {
	'seed': 22
});
// returns [ ~-0.643, ~0.937, ~0.049 ]

var out = random( 3, {
    'seed': 22
});
// returns [ ~-0.643, ~0.937, ~0.049 ]

If no seed option is supplied, each function call uses a common underlying uniform number generator. A positive-integer seed for this underlying generator can be supplied by setting the seed property of the exported function.

random.seed = 11;
var out = random();
// returns ~-0.921

var out = random();
// returns ~0.389

random.seed = 11;
var out = random();
// returns ~-0.921

var out = random();
// returns ~0.389

By default, the output data structure is a generic array. To output a typed array or matrix, set the dtype option.

var out;

out = random( 5, {
	'dtype': 'float32'
});
// returns Float32Array( [~0.166,~0.916,~-0.003,-0.08,~-2.608] )

out = random( [3,2], {
	'dtype': 'float64'
});
/*
	[ ~-0.482 ~0.274
	  ~0.725 ~1.113
	  ~0.608 1.05 ]
*/

Notes:

Currently, for more than 2 dimensions, the function outputs a generic array and ignores any specified dtype.

var out = random( [2,1,3], {
	'dtype': 'float32'
});
// returns [ [ [~0.873,~-0.510,~-0.370] ], [ [~0.393,~-0.233,~0.907] ] ]

Method

The used algorithm to generate normal random variables is the "Improved Ziggurat method" developed by J. Doornik. In a speed comparison of different algorithms, Doornik found that the Ziggurat method was two or three times faster than the commonly used polar method (Box-Mueller Transform) when generating 10^9 standard normal random numbers.

Reference:

Doornik, J. a. (2005). An Improved Ziggurat Method to Generate Normal Random Samples.

Examples

var random = require( 'distributions-normal-random' ),
	out;

// Set seed
random.seed = 11;

// Plain arrays...

// 1x10:
out = random( 10 );

// 2x1x3:
out = random( [2,1,3] );

// 5x5x5:
out = random( [5,5,5] );

// 10x5x10x20:
out = random( [10,5,10,20] );

// Typed arrays...
out = random( 10, {
	'dtype': 'float32'
});

// Matrices...
out = random( [3,2], {
	'dtype': 'float64'
});

To run the example code from the top-level application directory,

$ node ./examples/index.js

Tests

Unit

Unit tests use the Mocha test framework with Chai assertions. To run the tests, execute the following command in the top-level application directory:

$ make test

All new feature development should have corresponding unit tests to validate correct functionality.

Test Coverage

This repository uses Istanbul as its code coverage tool. To generate a test coverage report, execute the following command in the top-level application directory:

$ make test-cov

Istanbul creates a ./reports/coverage directory. To access an HTML version of the report,

$ make view-cov

License

MIT license.

distributions-normal-random v0.0.1