Branch data     Line data    Source code

       1                 :            : #pragma once
       2                 :            : 
       3                 :            : #include <Eigen/Dense>
       4                 :            : #include <mutable/mutable.hpp>
       5                 :            : #include <vector>
       6                 :            : 
       7                 :            : 
       8                 :            : namespace m {
       9                 :            : 
      10                 :            : using MatrixRXf = Eigen::Matrix<float, Eigen::Dynamic, Eigen::Dynamic, Eigen::RowMajor>;
      11                 :            : 
      12                 :            : /** Compute initial cluster centroids using *k*-means++ algorithm.  See https://en.wikipedia.org/wiki/K-means%2B%2B.
      13                 :            :  *
      14                 :            :  * @param data the data to cluster, as `Eigen::Matrix`; rows are data points, columns are attributes of data points
      15                 :            :  * @param k    the number of initial centroids to compute
      16                 :            :  * @return     a `MatrixRXf` of `k` data points (rows) chosen as the `k` initial centroids for *k*-means
      17                 :            :  */
      18                 :            : MatrixRXf M_EXPORT kmeans_plus_plus(const Eigen::MatrixXf &data, unsigned k);
      19                 :            : 
      20                 :            : /** Clusters the given data according to the *k*-means algorithm.
      21                 :            :  *
      22                 :            :  * The `k` clusters are represented by `k` unique integers (usually from the range *[0, k)*).  The *k*-means algorithm
      23                 :            :  * assigns an integer label to each data point, identifying to which cluster the data point is assigned.  See
      24                 :            :  * https://en.wikipedia.org/wiki/K-means_clustering.
      25                 :            :  *
      26                 :            :  * @param data the data to cluster, as `Eigen::Matrix`; rows are data points, columns are attributes of data points
      27                 :            :  * @param k    the number of clusters to form (more like an upper bound, as clusters may be empty)
      28                 :            :  * @return     a `std::pair` of a `std::vector<unsigned>` assigning a label to each data point and an `Eigen::Matrix` of
      29                 :            :  *             `k` rows with the centroids of the formed clusters
      30                 :            :  */
      31                 :            : std::pair<std::vector<unsigned>, MatrixRXf> M_EXPORT kmeans_with_centroids(const Eigen::MatrixXf &data, unsigned k);
      32                 :            : 
      33                 :            : /** Clusters the given data according to the *k*-means algorithm.
      34                 :            :  *
      35                 :            :  * The `k` clusters are represented by `k` unique integers (usually from the range *[0, k)*).  The *k*-means algorithm
      36                 :            :  * assigns an integer label to each data point, identifying to which cluster the data point is assigned.  See
      37                 :            :  * https://en.wikipedia.org/wiki/K-means_clustering.
      38                 :            :  *
      39                 :            :  * @param data the data to cluster, as `Eigen::Matrix`; rows are data points, columns are attributes of data points
      40                 :            :  * @param k    the number of clusters to form (more like an upper bound, as clusters may be empty)
      41                 :            :  * @return     a `std::vector<unsigned>` assigning a label to each data point
      42                 :            :  */
      43                 :          5 : inline std::vector<unsigned> M_EXPORT kmeans(const Eigen::MatrixXf &data, unsigned k) {
      44                 :          5 :     return kmeans_with_centroids(data, k).first;
      45                 :            : }
      46                 :            : 
      47                 :            : }