MOYSES

Description

MOYSES is a Support Vector Machine (SVM) library for node.js using TypeScript. It's used for binary classification purposes using n-dimensional datasets.

Overview

Directory structure

.
├── core
│   ├── engine
│   │   └── svm.ts
│   └── kernels
│       └── kernels.ts
├── index.ts
├── types
│   └── dataset_type.ts
└── utils
    ├── dataset_generation
    │   ├── dataset_generator.ts
    │   └── generate_points.ts
    └── utils.ts
    

How to build the library to be used in production-ready projects?

npm install moyses

How to use?

import * as Moyses from 'moyses'
 
// will generate 10 pairs of labeled data you might as well wanna use your own dataset
const dataset: Moyses.IDataset = new Moyses.DatasetGenerator('CIRCULAR',10).generate();
 
//instanciate SVM
const svm = new Moyses.SVM(dataset, 5, 'RBF', 15 );
 
//classify data
const positiveResult = svm.predict([0,0]);
const negativeResult = svm.predict([50,50]);
 

Svm arguments :

• dataset: type: IDataset Interface can be found in lib/types/dataset_type.ts or see example below.
• c: type: number c parameter for soft margin classification.
• kernel: type: string Only 'RBF' kernel is supported yet.
• OPTIONAL rbfSigma: type: number variance. Default value = 15 .

DatasetGenerator arguments :

• shape: type: string Overall shape of dataset (CIRCULAR, LINEAR, XOR).
• total: type: number Total amount of data pairs (1 and -1 output).
• OPTIONAL dimension: type: number dataset dimension default is 2 dim.

Note: Dataset boundaries are fixed. This should be fixed at some point..

Example dataset :

const circularDataset: IDataset = {
  points: [
    [ 77.08537142627756, 60.7455136985482 ],
    [ 54.94324221651883, 63.78584077042318 ],
    [ 45.124087171506936, 80.97650097253724 ],
    [ 62.00480777917741, 49.642444449970675 ],
    [ 56.958382663885864, 81.27710664286386 ],
    [ 52.72767259658451, 66.03517399586579 ],
    [ 19.518515661340157, 35.12014495118882 ],
    [ 58.87894639269981, 59.27927960679746 ],
    [ 13.59822313333904, 61.66342807818599 ],
    [ 37.01348768362775, 54.679365456721584 ],
    [ 85.01654232561876, 46.57532675823407 ],
    [ 34.70627848361286, 44.84248665899513 ],
    [ 63.443893468418494, 74.07028656564599 ],
    [ 61.456705623249455, 41.09439124577563 ],
    [ 84.26782294646438, 26.269714017498337 ],
    [ 37.44407046741475, 50.98956479733988 ],
    [ 37.53801531593166, 79.73505569185346 ],
    [ 61.308207468398585, 44.41090753575729 ],
    [ 49.57073028314457, 5.715350047914129 ],
    [ 63.640430592148775, 39.56876863124383 ]
  ],
  labels: [
     1, -1,  1, -1,  1, -1,  1,
    -1,  1, -1,  1, -1,  1, -1,
     1, -1,  1, -1,  1, -1
  ]
}

NPM custom commands

• build: Build the JavaScript files.
• build:watch: Build the JavaScript files in watch mode.
• test: Run jest in test mode.
• test:watch: Run jest in interactive test mode.
• docs: Generate the docs directory.
• lint: Runs linter on the whole project.

Other/Optional considerations

The model converges, however it is a simplified version of the sequential minimum optimisation algorithm published by John C.Platt.

Please follow the links below for more informations on the model.

• https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/tr-98-14.pdf
• http://cs229.stanford.edu/materials/smo.pdf

License

Bastien GUIHARD