What is this?

A simple implementation of the k-means clustering algorithm for C++.

Installation

This is a header-only library, simply #include "kmeans.hpp" from core/include/ and you're good to go!

Usage

Create

using namespace cluster;
typedef Vector<float, 2> Vector2f;

const std::vector<Vector2f> observations = ...;
const size_t cluster_count = 3;
kMeans<float, 2, distance::euclidean2<float, 2>> solver(

    cluster_count,
    &observations[0], observations.size()
); // for two dimensional floating-point data and
   // euclidean distance measure.

Initialize means

const std::vector<Vector2f> initial_means = ...;
solver.initialize(&initial_means[0]);

Run the algorithm

Automatically

const size_t max_iterations = 3;
solver.run(max_iterations);

Manually

const size_t max_iterations = 3;
size_t i = 0;
while (i < max_iterations && solver.assign())
{
    solver.update();
    ++ i;
}

Inspect means

Vector2f cluster_mean = solver.mean(i);
// Where i in range [0, cluster_count)

Further documentation

Refer to the docs.

Demo

An example application can be found under demo/. The python script under demo/scripts/ was used to generate some random data points, cluster them using the demo executable, and export visualization of intermediate steps of the algorithm under demo/scripts/visuals/.

License

This software is licensed under the MIT License. See the LICENSE file for more information.