kmeans_clustering.hpp

Go to the documentation of this file.

/* Copyright (c) 2025 Otto Link. Distributed under the terms of the GNU General
   Public License. The full license is in the file LICENSE, distributed with
   this software. */
#pragma once
#include "dkm.hpp" // https://github.com/genbattle/dkm
 
#include "point_sampler/internal/nanoflann_adaptator.hpp"
#include "point_sampler/point.hpp"
 
namespace ps
{
 
template <typename T, size_t N>
std::pair<std::vector<Point<T, N>>, std::vector<size_t>> kmeans_clustering(
    const std::vector<Point<T, N>> &points,
    size_t                          k_clusters,
    bool                            normalize_data = true,
    size_t                          max_iterations = 100)
{
  using DKMPoint = std::array<T, N>;
  std::vector<DKMPoint> data;
  data.reserve(points.size());
 
  if (normalize_data)
  {
    auto points_normalized = points;
    normalize_points(points_normalized);
 
    for (const auto &p : points_normalized)
      data.push_back(p.coords);
  }
  else
  {
    for (const auto &p : points)
      data.push_back(p.coords);
  }
 
  auto result = dkm::kmeans_lloyd<T>(data, k_clusters, max_iterations);
 
  auto &centroids_dkm = std::get<0>(result);
  auto &labels_dkm_u32 = std::get<1>(result); // uint32_t labels from DKM
 
  // Convert centroids
  std::vector<Point<T, N>> centroids;
  centroids.reserve(centroids_dkm.size());
  for (const auto &c : centroids_dkm)
    centroids.emplace_back(c);
 
  // Convert labels to size_t
  std::vector<size_t> labels;
  labels.reserve(labels_dkm_u32.size());
  for (auto v : labels_dkm_u32)
    labels.push_back(static_cast<size_t>(v));
 
  return {std::move(centroids), std::move(labels)};
}
 
} // namespace ps