node-svm-smo

Support Vector Machine implementing Sequential Minimal Optimisation algorithm for Node.js

Installation

npm i node-svm-smo

Basic usage

import the module : import SVM from "node-svm-smo";

Example

import SVM from "node-svm-smo";

const data = [
  [63.94467417052296,81.0217272621887],
  [54.87300013454443,60.84061766333669],
  [45.52048694403639,81.12464276459878],
  [48.965524328511705,57.18776467921587],
  [74.52679297504449,83.83083375121797],
  [54.37811765569671,56.03892121990233],
  [44.26860038064747,19.667730330959163],
  [49.174828952241675,45.63294964476442],
  [68.38409733522553,71.71723595963316],
  [59.78161085912989,61.55507106057365],
  [6.6500878665167775,39.49294636329541],
  [44.48056808348252,48.662212578138195]
];

const labels = [-1,1,-1,1,-1,1,-1,1,-1,1,-1,1];

const C = 5;
const kernel = 'RBF';
const RBFSigma = 15;

const MySVM = new SVM(data,labels,C,kernel,RBFSigma);

MySVM.predict([50,50]);
// expected output : 1

check out https://github.com/guihardbastien/Support-Vector-Machine for data visualisation.

Parameters

• data : Array of feature vectors should be of form :

const data = [[X1,Y1,...,Z1],[X2,Y2,...,Z2],...,[Xx,Yx,...,Zx]]

• labels: Single array of data output, should match data.length

• C : soft margin parameter

• kernel : kernel used for the the kernel trick

  • RBF: Radial Basis Function

Options

• RBFSigma:

Methods

• predict(data) : returns binary classification of an input

TO DO

• [ ] Add default parameters
• [ ] Add more kernels
• [ ] Write tests
• [ ] Comment code :trollface:

Sources

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

License