mokolo
Collection of machine learning algorithms: Non-Negative Matrix Factorization for JavaScript
npm install mokolo
Want to see pretty graphs? Log in now!
2 | downloads in the last month |
Last Published By | |
---|---|
Version | 0.0.2 last updated 2 years ago |
License | MIT |
Keywords | vector, matrix, geometry, math, features, extraction |
Homepage | https://github.com/gustii/mokolo |
Dependencies | sylvester |
mokolo
mokolo intends to become a collection of machine learning algorithms for Node.js. The current release supports only the Non-Negative Matrix Factorization (NMF) algorithm; more are coming soon. Feedbacks and contributions are greatly appreciated.
Install
The easiest way to install mokolo is through npm, the nodejs package manager.
npm install mokolo
Non-Negative Matrix Factorization (NMF)
NMF is an advanced technique which allows breaking down a set of numerical observations into their component parts. It can be used in numerous fields including text mining, spectral data analysis, scalable Internet distance prediction, non-stationary speech denoising and bioinformatics. Learn more on Wikipedia.
Prerequisite
NMF uses extensively Sylvester, a vector and matrix math library for JavaScript written by James Coglan. If Sylvester is not already installed, npm will do it for you during mokolo's installation.
Initialization
var mokolo = require('mokolo'),
nmf = new mokolo.NMF();
Usage
nmf.factorize(options, callback);
options
options is a hash with the following structure.
{
matrix : M // matrix to be factorized; e.g. a matrix object generated from an array using Sylvester
,features : number // the dimensions (number x number) of the features matrix to be computed; e.g. 2
,iterations : number // the number of iterations; e.g. 1000
,precision : number // the position at which the `diff` will be rounded; e.g. 1e-15
}
callback
callback is a function taking six (6) parameters
callback(W, H, WH, diff, iter, precision){
}
W
- Computed Weights matrrixH
- Computed Features matrixWH
- multiplication of computed matrices W and Hdiff
- Difference betweenM
andWH
iter
- number of iterations needed to getWH
close enough toM
precision
- same as above
Example
Given the following dataset represented by an array of arrays
var dArray;
dArray = [
[29, 29],
[43, 33],
[15, 25],
[40, 28],
[24, 11],
[29, 29],
[37, 23],
[21, 6]
];
Using Sylvester a matrix can be generated from the array above.
var $M = require('sylvester').Matrix.create;
var D = $M(dArray);
Now factorization can start.
nmf.factorize({
matrix: D
,features: 2
,iterations: 1000
,precision: 1e-10
},
function(W, H, WH, diff, iter, precision){
console.log('Weights matrix');
console.log(W)
console.log('Features matrix');
console.log(H);
console.log('Computed matrix');
console.log(WH);
console.log(iter);
console.log(diff);
}
);
TO DO
- Set default values to NMF parameters
- More test coverage
- Support for more algorithms; e.g. clustering, K-nearest neighbors, etc.
License
Copyright (c) 2012, Gustave Nganso
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.