PartitionalClustering.h

Go to the documentation of this file.

/**
 *  \file IMP/statistics/PartitionalClustering.h
 *  \brief Compute a distance metric between two points
 *
 *  Copyright 2007-2014 IMP Inventors. All rights reserved.
 *
 */

#ifndef IMPSTATISTICS_PARTITIONAL_CLUSTERING_H
#define IMPSTATISTICS_PARTITIONAL_CLUSTERING_H

#include <IMP/statistics/statistics_config.h>
#include <IMP/macros.h>
#include <IMP/base_types.h>
#include <IMP/base/Object.h>

IMPSTATISTICS_BEGIN_NAMESPACE

//! A base class for clustering results where each item is in one cluster
/** Data items are represented by an index. In a partitional clustering
    each item can only be contained on one cluster. The indexes used refer
    to the index used in the Metric or Embedding used to generate the
    clustering.
*/
class IMPSTATISTICSEXPORT PartitionalClustering : public IMP::base::Object {
 public:
  PartitionalClustering(std::string name) : Object(name) {}
  virtual unsigned int get_number_of_clusters() const = 0;
  /** Return the list of indexes in the ith cluster.
   */
  virtual const Ints& get_cluster(unsigned int cluster_index) const = 0;
  /** Return the index of an item that "typifies" the ith cluster.
   */
  virtual int get_cluster_representative(unsigned int cluster_index) const = 0;
  IMP_REF_COUNTED_NONTRIVIAL_DESTRUCTOR(PartitionalClustering);
};

/** Check that the clustering is a valid clustering of n elements. An
 exception is thrown if it is not, if the build is not a fast build.*/
IMPSTATISTICSEXPORT void validate_partitional_clustering(
    PartitionalClustering* pc, unsigned int n);

IMPSTATISTICS_END_NAMESPACE
#endif /* IMPSTATISTICS_PARTITIONAL_CLUSTERING_H */