121 lines
3.1 KiB
C++
121 lines
3.1 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
|
||
|
|
----
|
||
|
|
|
||
|
|
//paragraph [1] Title: [{\Large \bf \begin{center}] [\end{center}}]
|
||
|
|
//[TOC] [\tableofcontents]
|
||
|
|
|
||
|
|
[1] Implementation of the CompareKmeans algorithm
|
||
|
|
|
||
|
|
|
||
|
|
1 Implementation of the CompareKmeans algorithm
|
||
|
|
|
||
|
|
*/
|
||
|
|
|
||
|
|
/* Authors: Greg Hamerly and Jonathan Drake
|
||
|
|
* Feedback: hamerly@cs.baylor.edu
|
||
|
|
* See: http://cs.baylor.edu/~hamerly/software/kmeans.php
|
||
|
|
* Copyright 2014
|
||
|
|
*/
|
||
|
|
|
||
|
|
#include "compare_kmeans.h"
|
||
|
|
#include "general_functions.h"
|
||
|
|
#include <cassert>
|
||
|
|
#include <cmath>
|
||
|
|
#include <algorithm>
|
||
|
|
|
||
|
|
void CompareKmeans::free() {
|
||
|
|
OriginalSpaceKmeans::free();
|
||
|
|
delete [] centersDist2div4;
|
||
|
|
centersDist2div4 = NULL;
|
||
|
|
}
|
||
|
|
|
||
|
|
void CompareKmeans::update_center_dists(int threadId) {
|
||
|
|
// find the inter-center distances
|
||
|
|
for (int c1 = 0; c1 < k; ++c1) {
|
||
|
|
#ifdef USE_THREADS
|
||
|
|
if (c1 % numThreads != threadId) {
|
||
|
|
continue;
|
||
|
|
}
|
||
|
|
#endif
|
||
|
|
|
||
|
|
centersDist2div4[c1 * k + c1]
|
||
|
|
= std::numeric_limits<double>::max();
|
||
|
|
|
||
|
|
for (int c2 = c1 + 1; c2 < k; ++c2) {
|
||
|
|
centersDist2div4[c1 * k + c2] = centersDist2div4[c2
|
||
|
|
* k + c1] = centerCenterDist2(c1, c2) / 4.0;
|
||
|
|
}
|
||
|
|
}
|
||
|
|
}
|
||
|
|
|
||
|
|
void CompareKmeans::initialize(Dataset const *aX, unsigned short aK,
|
||
|
|
unsigned short *initialAssignment, int aNumThreads) {
|
||
|
|
OriginalSpaceKmeans::initialize(aX, aK, initialAssignment,
|
||
|
|
aNumThreads);
|
||
|
|
centersDist2div4 = new double[k * k];
|
||
|
|
std::fill(centersDist2div4, centersDist2div4 + k * k, 0.0);
|
||
|
|
}
|
||
|
|
|
||
|
|
int CompareKmeans::runThread(int threadId, int maxIterations) {
|
||
|
|
int iterations = 0;
|
||
|
|
|
||
|
|
int startNdx = start(threadId);
|
||
|
|
int endNdx = end(threadId);
|
||
|
|
|
||
|
|
while ((iterations < maxIterations) && ! converged) {
|
||
|
|
++iterations;
|
||
|
|
|
||
|
|
update_center_dists(threadId);
|
||
|
|
synchronizeAllThreads();
|
||
|
|
|
||
|
|
for (int i = startNdx; i < endNdx; ++i) {
|
||
|
|
int minClass = assignment[i];
|
||
|
|
double minDist2 = pointCenterDist2(i, minClass);
|
||
|
|
|
||
|
|
for (int j = 0; j < k; ++j) {
|
||
|
|
// center-center squared distances are already divided by 4.0
|
||
|
|
if (centersDist2div4[j * k + minClass] > minDist2)
|
||
|
|
continue;
|
||
|
|
|
||
|
|
if (j == minClass) continue;
|
||
|
|
|
||
|
|
const double dist2 = pointCenterDist2(i, j);
|
||
|
|
|
||
|
|
if (dist2 < minDist2) {
|
||
|
|
minDist2 = dist2;
|
||
|
|
minClass = j;
|
||
|
|
} else if (dist2 == minDist2) {
|
||
|
|
if (j < minClass) {
|
||
|
|
minClass = j;
|
||
|
|
}
|
||
|
|
}
|
||
|
|
}
|
||
|
|
|
||
|
|
if (assignment[i] != minClass) {
|
||
|
|
changeAssignment(i, minClass, threadId);
|
||
|
|
}
|
||
|
|
}
|
||
|
|
|
||
|
|
verifyAssignment(iterations, startNdx, endNdx);
|
||
|
|
|
||
|
|
synchronizeAllThreads();
|
||
|
|
if (threadId == 0) {
|
||
|
|
int furthestMovingCenter = move_centers();
|
||
|
|
|
||
|
|
converged = (0.0 == centerMovement[furthestMovingCenter]);
|
||
|
|
}
|
||
|
|
|
||
|
|
synchronizeAllThreads();
|
||
|
|
}
|
||
|
|
|
||
|
|
return iterations;
|
||
|
|
}
|
||
|
|
|
||
|
|
|