/home/runner/work/deep-learning-algo-impls/deep-learning-algo-impls/include/ml/kmeans.hpp

A class implementing K-Means clustering algorithm.

A class implementing K-Means clustering algorithmThis class provides functionality for unsupervised clustering using the K-Means algorithm. It partitions data into k clusters by minimizing within-cluster sum of squares.

Template Parameters

T	The data type for matrix elements (typically float or double)

// Create a K-Means object with 3 clusters
KMeans<double> kmeans(3);
 
// Fit K-Means to data
Matrix<double> data = your_data_matrix;
kmeans.fit(data);
 
// Predict cluster labels for new data
std::vector<int> labels = kmeans.predict(data);

#pragma once
 
#include <limits>
#include <random>
#include <vector>
#include "../utils/matrix.hpp"
 
namespace ml {
    using namespace utils;
 
    template<typename T>
    class KMeans {
    public:
        KMeans(size_t k, size_t max_iters = 300, T tol = 1e-4, int random_state = -1);
 
        void fit(const Matrix<T> &data);
 
        std::vector<int> predict(const Matrix<T> &data) const;
 
        std::vector<int> fit_predict(const Matrix<T> &data);
 
        Matrix<T> cluster_centers() const;
 
        T inertia() const;
 
        size_t n_iter() const;
 
    private:
        size_t k_; 
        size_t max_iters_; 
        T tol_; 
        int random_state_; 
        Matrix<T> centroids_; 
        T inertia_; 
        size_t n_iter_; 
        bool is_fitted_; 
 
        void init_centroids(const Matrix<T> &data);
 
        std::vector<int> assign_clusters(const Matrix<T> &data) const;
 
        void update_centroids(const Matrix<T> &data, const std::vector<int> &labels);
 
        T squared_distance(const std::vector<T> &point1, const std::vector<T> &point2) const;
    };
 
    using KMeansF = KMeans<float>;
    using KMeansD = KMeans<double>;
 
} // namespace ml