@softnami/softmaxregression NPM

Softmax Regression

Author: Hussain Mir Ali

This module contains logic to run the Softmax Regression algorithm.

External Libraries Used:

math License: https://github.com/josdejong/math/blob/master/LICENSE
mocha License: https://github.com/mochajs/mocha/blob/master/LICENSE
sinon Licencse: https://github.com/sinonjs/sinon/blob/master/LICENSE
yuidocjs License: https://github.com/yui/yuidoc/blob/master/LICENSE
nodeJS License: https://github.com/nodejs/node/blob/master/LICENSE

Installation:

run 'npm i @softnami/softmaxregression'

Sample usage:

import {SoftmaxRegression} from '@softnami/softmaxregression';
import * as math from 'mathjs';

const callback = (data)=> {
    console.log(data);
};

const generateData = ()=>{
    let sample_y = []; //one-hot encoded classes: [1,2,3] -> [[1, 0,0],[0, 1,0],[0, 0,1]]
    let y = [], num = 0, 
    x = (math.random(math.matrix([1000, 78]), -1, 1));

    for(let i =0 ; i< 100; i++){
      sample_y[i] = [];
      for(let j =0; j< 12; j++){
        if(i==j){
          sample_y[i][j]= 1;
        }
        else{
          sample_y[i][j]= 0;
        }
      }  
    }

    for (let i = 0; i < 1000; i++) {
      num = Math.floor((Math.random() * 100) + 0);
      y.push(sample_y[num]);
    }

    return [x, y];
};

const softmaxRegression = new SoftmaxRegression({
    'notify_count': 1,
    'momentum': 0.5,
    'parameter_size': [78, 12], //[number of  input features, total number of  output classes]
    'max_epochs': 100,
    'weight_initialization_range': [-0.5, 0.5],
    'threshold': 0.1,
    'batch_size': 256,
    'iteration_callback': callback,
    'learningRate': 0.009,
    'regularization_parameter': Math.exp(-4)
  });

let [x, y] = generateData();

console.log('Start training.');
softmaxRegression.startRegression(x, math.matrix(y));

Testing:

For unit testing Mocha and Sinon have been used.
Run 'npm test', if timeout occurs then increase timeout in test script.

Documentation

The documentation is available in the 'out' folder of this project. Open the 'index.html' file under the 'out' folder with Crhome or Firefox.
To generate the documentation install 'yuidocjs' globally then run 'yuidoc .' command in the main directory of this project.