SeregPie/KMeans

Implementation of the k-means algorithm to partition the values into the clusters.

KMeans

KMeans(values, means, {
  distance(value, otherValue) { /* euclidean distance */ },
  map(value) { /* identity */ },
  maxIterations: 1024,
  mean(...values) { /* centroid */ },
  random: Math.random,
})

Implementation of the k-means algorithm to partition the values into the clusters.

argument	description
values	An iterable of the values to be clustered.
means	Either an iterable of the initial means or the number of the clusters.
distance	A function to calculate the distance between two values.
map	A function to map the values.
maxIterations	The maximum number of iterations until the convergence.
mean	A function to calculate the mean value.
random	A function as the pseudo-random number generator.

Returns the clustered values as an array of arrays.

dependencies

• JustMyLuck
• VectorMath

setup

npm

npm install @seregpie/k-means

---

Import inside an ES module.

import KMeans from '@seregpie/k-means';

or

Import inside a CommonJS module.

let KMeans = require('@seregpie/k-means');

browser

<script src="https://unpkg.com/just-my-luck"></script>
<script src="https://unpkg.com/@seregpie/vector-math"></script>
<script src="https://unpkg.com/@seregpie/k-means"></script>

The module is globally available as KMeans.

usage

Let the initial means be chosen randomly.

let vectors = [[1, 4], [6, 2], [0, 4], [1, 3], [5, 1], [4, 0]];
let clusters = KMeans(vectors, 3);
// => [[[1, 4], [0, 4]], [[6, 2], [5, 1], [4, 0]], [[1, 3]]]

---

Provide the initial means.

let vectors = [[1, 4], [6, 2], [0, 4], [1, 3], [5, 1], [4, 0]];
let centroids = [[0, 7], [7, 0]];
let clusters = KMeans(vectors, centroids);
// => [[[1, 4], [0, 4], [1, 3]], [[6, 2], [5, 1], [4, 0]]]

---

Provide a map function to convert a value to a vector.

let Athlete = class {
  constructor(name, height, weight) {
    this.name = name;
    this.height = height;
    this.weight = weight;
  }
  toJSON() {
    return this.name;
  }
};
let athletes = [
  new Athlete('A', 185, 72), new Athlete('B', 183, 84), new Athlete('C', 168, 60),
  new Athlete('D', 179, 68), new Athlete('E', 182, 72), new Athlete('F', 188, 77),
  new Athlete('G', 180, 71), new Athlete('H', 180, 70), new Athlete('I', 170, 56),
  new Athlete('J', 180, 88), new Athlete('K', 180, 67), new Athlete('L', 177, 76),
];
let clusteredAthletes = KMeansPlusPlus(athletes, [athletes[0], athletes[1]], {
  map: athlete => [athlete.weight / athlete.height],
});
console.log(JSON.parse(JSON.stringify(clusteredAthletes)));
// => [['A', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'K'], ['B', 'J', 'L']]