ml-ransac

Fit a model to noisy data by excluding outliers. This is an implementation of the RANSAC algorithm..

Installation

$ npm i ml-ransac

Examples

Finding the best linear regression which doesn't take the outliers into account:

import { levenbergMarquardt } from 'ml-levenberg-marquardt'; // library allowing you to find a regression of any type
import { ransac } from 'ml-ransac';

// defining the model function: a line
export type LineFunction = (x: number) => number;

export function line([a, b]: number[]): LineFunction {
  return (x: number) => a * x + b;
}

// defining the fit function: we create a linear regression using levenberg-maquard
export interface LinearRegressionOptions {
  /**
   * Initial parameter values.
   *
   * @default [0,0]
   */
  initialValues?: number[];
}

export function linearRegression(
  source: number[],
  destination: number[],
  options: LinearRegressionOptions = {},
): number[] {
  const { initialValues = [0, 0] } = options;
  const result = levenbergMarquardt({ x: source, y: destination }, line, {
    initialValues,
  });

  return result.parameterValues;
}

// data
const source = [-2, -1, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10];
const destination = [-6, -2, 0, 5, 0, 0, -1, 1, 0, 0, -1, 0, 3];

// finding the best model
const result = ransac(source, destination, {
  distanceFunction,
  fitFunction,
  modelFunction,
  threshold: 1.1,
  sampleSize: 2,
  maxNbIterations: 10,
  seed: 0,
});

// result is :
//   {
//     nbIterations: 10,
//     modelParameters: [0, 0],
//     inliers: [2, 4, 5, 6, 7, 8, 9, 10, 11],
//     error: 1,
//   }

License

MIT