79 lines
2.4 KiB
C
79 lines
2.4 KiB
C
|
/*
|
||
|
|
----
|
||
|
|
This file is NOT part of SECONDO.
|
||
|
|
* Authors: Greg Hamerly and Jonathan Drake
|
||
|
|
* Feedback: hamerly@cs.baylor.edu
|
||
|
|
* See: http://cs.baylor.edu/~hamerly/software/kmeans.php
|
||
|
|
* Copyright 2014
|
||
|
|
*
|
||
|
|
* A kernelized version of Elkan's algorithm. It would be nice to make this
|
||
|
|
* inherit from both KernelKmeans and ElkanKmeans, but for now we basically
|
||
|
|
* replicate code.
|
||
|
|
----
|
||
|
|
|
||
|
|
//paragraph [1] Title: [{\Large \bf \begin{center}] [\end{center}}]
|
||
|
|
//[TOC] [\tableofcontents]
|
||
|
|
|
||
|
|
[1] Declarations for the Eklan's' algorithm
|
||
|
|
|
||
|
|
|
||
|
|
1 Declarations for the Elkan's algorithm
|
||
|
|
|
||
|
|
*/
|
||
|
|
|
||
|
|
#ifndef ELKAN_KERNEL_KMEANS_H
|
||
|
|
#define ELKAN_KERNEL_KMEANS_H
|
||
|
|
|
||
|
|
/* Authors: Greg Hamerly and Jonathan Drake
|
||
|
|
* Feedback: hamerly@cs.baylor.edu
|
||
|
|
* See: http://cs.baylor.edu/~hamerly/software/kmeans.php
|
||
|
|
* Copyright 2014
|
||
|
|
*
|
||
|
|
* A kernelized version of Elkan's algorithm. It would be nice to make this
|
||
|
|
* inherit from both KernelKmeans and ElkanKmeans, but for now we basically
|
||
|
|
* replicate code.
|
||
|
|
*/
|
||
|
|
|
||
|
|
#include "kernel_kmeans.h"
|
||
|
|
|
||
|
|
class ElkanKernelKmeans : public KernelKmeans {
|
||
|
|
public:
|
||
|
|
ElkanKernelKmeans(Kernel const *k) : KernelKmeans(k),
|
||
|
|
centerCenterDistDiv2(NULL), s(NULL), upper(NULL), lower(NULL) {}
|
||
|
|
virtual ~ElkanKernelKmeans() { free(); }
|
||
|
|
virtual std::string getName() const {
|
||
|
|
std::ostringstream out;
|
||
|
|
out << "elkan_kernel(" << kernel.getName() << ")";
|
||
|
|
return out.str();
|
||
|
|
}
|
||
|
|
virtual void free();
|
||
|
|
virtual void initialize(Dataset const *aX, unsigned short aK,
|
||
|
|
unsigned short *initialAssignment, int aNumThreads);
|
||
|
|
|
||
|
|
protected:
|
||
|
|
int runThread(int threadId, int maxIterations);
|
||
|
|
void update_bounds(int startNdx, int endNdx);
|
||
|
|
void update_center_dists(int threadId);
|
||
|
|
void computeCenterMovement(int threadId);
|
||
|
|
|
||
|
|
// Matrix in an array of distance between each center, divided by 2.
|
||
|
|
double *centerCenterDistDiv2;
|
||
|
|
|
||
|
|
// Distance between center j and its closest other center, divided by 2.
|
||
|
|
double *s;
|
||
|
|
|
||
|
|
// Upper bound for each point.
|
||
|
|
double *upper;
|
||
|
|
|
||
|
|
// Matrix in an array of k lower bounds for each point.
|
||
|
|
double *lower;
|
||
|
|
|
||
|
|
// These are used to hold the new memberships and new squared center
|
||
|
|
// norms (Cc means <c,c>) from one iteration to the next.
|
||
|
|
std::vector<std::vector<unsigned int> > newMemberships;
|
||
|
|
std::vector<double> newCc;
|
||
|
|
};
|
||
|
|
|
||
|
|
#endif
|
||
|
|
|