#667: PAM BUILD algorithm implementation (Python version).

CHANGES

@@ -6,6 +6,9 @@ CHANGE NOTES FOR 0.11.0 (STARTED Nov 26, 2020), (RELEASED: -)
 
 GENERAL CHANGES:
 
+- Introduced PAM BUILD algorithm to generate initial medoids (Python: `pyclustering.cluster.kmedoids`, C++: `pyclustering::clst::pam_build`).
+  See: https://github.com/annoviko/pyclustering/issues/667
+
 - Parallel optimization of K-Medoids (PAM) algorithm (Python: `pyclustering.cluster.kmedoids`, C++: `pyclustering::clst::kmedoids`).
   See: https://github.com/annoviko/pyclustering/issues/667
 

ccore/include/pyclustering/cluster/kmedoids_data.hpp

@@ -33,6 +33,8 @@ class kmedoids_data : public cluster_data {
 private:
     medoid_sequence     m_medoids = { };
 
+    std::size_t         m_iterations = 0;
+
 public:
     /**
     *
@@ -69,17 +71,31 @@ class kmedoids_data : public cluster_data {
 public:
     /**
     *
-    * @brief    Returns medoids that corresponds to allocated clusters.
+    * @brief    Returns reference medoids that corresponds to allocated clusters.
     *
     */
     medoid_sequence & medoids() { return m_medoids; }
 
     /**
     *
-    * @brief    Returns medoids that corresponds to allocated clusters.
+    * @brief    Returns constant reference to medoids that corresponds to allocated clusters.
     *
     */
     const medoid_sequence & medoids() const { return m_medoids; }
+
+    /*
+    
+    @brief      Returns reference to the amount of iterations that were performed during the clustering process.
+
+    */
+    std::size_t & iterations() { return m_iterations; }
+
+    /*
+
+    @brief      Returns the amount of iterations that were performed during the clustering process.
+
+    */
+    std::size_t iterations() const { return m_iterations; }
 };
 
 

ccore/include/pyclustering/interface/kmedoids_interface.h

@@ -23,6 +23,7 @@
 enum kmedoids_package_indexer {
     KMEDOIDS_PACKAGE_INDEX_CLUSTERS = 0,
     KMEDOIDS_PACKAGE_INDEX_MEDOIDS,
+    KMEDOIDS_PACKAGE_INDEX_ITERATIONS,
     KMEDOIDS_PACKAGE_SIZE
 };
 

ccore/src/cluster/kmedoids.cpp

@@ -82,7 +82,7 @@ void kmedoids::process(const dataset & p_data, const kmedoids_data_t p_type, kme
         current_deviation = update_clusters();
     }
 
-    for (std::size_t iteration = 0; (iteration < m_itermax) && (changes > m_tolerance); iteration++) {
+    for (p_result.iterations() = 0; (p_result.iterations() < m_itermax) && (changes > m_tolerance); p_result.iterations()++) {
         const double swap_cost = swap_medoids();
 
         if (swap_cost != NOTHING_TO_SWAP) {

ccore/src/interface/kmedoids_interface.cpp

@@ -45,6 +45,9 @@ pyclustering_package * kmedoids_algorithm(const pyclustering_package * const p_s
     ((pyclustering_package **) package->data)[KMEDOIDS_PACKAGE_INDEX_CLUSTERS] = create_package(&output_result.clusters());
     ((pyclustering_package **) package->data)[KMEDOIDS_PACKAGE_INDEX_MEDOIDS] = create_package(&output_result.medoids());
 
+    std::vector<std::size_t> iteration_storage(1, output_result.iterations());
+    ((pyclustering_package **)package->data)[KMEDOIDS_PACKAGE_INDEX_ITERATIONS] = create_package(&iteration_storage);
+
     return package;
 }
 

pyclustering/cluster/center_initializer.py

@@ -166,7 +166,7 @@ def __init__(self, data, amount_centers, amount_candidates=None, **kwargs):
         """!
         @brief Creates K-Means++ center initializer instance.
         
-        @param[in] data (array_like): List of points where each point is represented by list of coordinates.
+        @param[in] data (array_like): Points where each point is represented by list of coordinates.
         @param[in] amount_centers (uint): Amount of centers that should be initialized.
         @param[in] amount_candidates (uint): Amount of candidates that is considered as a center, if the farthest points
                     (with the highest probability) should be considered as centers then special constant should be used

pyclustering/cluster/examples/kmedoids_examples.py

@@ -27,10 +27,13 @@ def template_clustering(start_medoids, path, tolerance=0.25, show=True, **kwargs
         sample = calculate_distance_matrix(sample)
 
     metric = distance_metric(type_metric.EUCLIDEAN_SQUARE, data=sample)
+
     kmedoids_instance = kmedoids(sample, start_medoids, tolerance, metric=metric, ccore=ccore, data_type=data_type)
     (ticks, result) = timedcall(kmedoids_instance.process)
 
     clusters = kmedoids_instance.get_clusters()
+    print("Iterations:", kmedoids_instance.get_iterations())
+    print([len(cluster) for cluster in clusters])
     print(clusters)
     medoids = kmedoids_instance.get_medoids()
     print("Sample: ", path, "\t\tExecution time: ", ticks, "\n")