/*
* This file is part of the LIRE project: http://lire-project.net
* LIRE is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* LIRE is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with LIRE; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*
* We kindly ask you to refer the any or one of the following publications in
* any publication mentioning or employing Lire:
*
* Lux Mathias, Savvas A. Chatzichristofis. Lire: Lucene Image Retrieval –
* An Extensible Java CBIR Library. In proceedings of the 16th ACM International
* Conference on Multimedia, pp. 1085-1088, Vancouver, Canada, 2008
* URL: http://doi.acm.org/10.1145/1459359.1459577
*
* Lux Mathias. Content Based Image Retrieval with LIRE. In proceedings of the
* 19th ACM International Conference on Multimedia, pp. 735-738, Scottsdale,
* Arizona, USA, 2011
* URL: http://dl.acm.org/citation.cfm?id=2072432
*
* Mathias Lux, Oge Marques. Visual Information Retrieval using Java and LIRE
* Morgan & Claypool, 2013
* URL: http://www.morganclaypool.com/doi/abs/10.2200/S00468ED1V01Y201301ICR025
*
* Copyright statement:
* --------------------
* (c) 2002-2013 by Mathias Lux (mathias@juggle.at)
* http://www.semanticmetadata.net/lire, http://www.lire-project.net
*/
package net.semanticmetadata.lire.utils.cv;
import net.semanticmetadata.lire.utils.MetricsUtils;
import java.util.*;
/**
* Simple, re-usable and straight k-means implementation based on double[] feature vectors and L2 distance.
* User: mlux
* Date: 20.09.13
* Time: 10:56
*/
public class KMeans {
List<double[]> features;
Cluster[] clusters;
public KMeans(List<double[]> featureList, int numberOfClusters) {
features = new ArrayList<double[]>(featureList);
clusters = new Cluster[numberOfClusters];
HashSet<double[]> means = new HashSet<double[]>();
while (means.size() < Math.min(numberOfClusters, featureList.size() / 2)) {
double[] e = features.get((int) Math.floor(Math.random() * features.size()));
double[] tmp = new double[e.length];
System.arraycopy(e, 0, tmp, 0, e.length);
means.add(tmp);
}
// init cluster centers.
Iterator<double[]> iterator = means.iterator();
for (int i = 0; i < clusters.length; i++) {
clusters[i] = new Cluster(iterator.next());
}
}
public double step() {
// init clusters:
for (int i = 0; i < clusters.length; i++) {
clusters[i].clearMembers();
assert (clusters[i].members.size() == 0);
}
// assign to new clusters:
for (int id = 0; id < features.size(); id++) {
double tmpDistance = Double.MAX_VALUE;
int currentCluster = -1;
for (int i = 0; i < clusters.length; i++) {
double distance = clusters[i].getDistance(features.get(id));
assert (distance >= 0);
if (distance < tmpDistance) {
tmpDistance = distance;
currentCluster = i;
}
}
clusters[currentCluster].addMember(id);
}
// recompute means:
for (int i = 0; i < clusters.length; i++) {
clusters[i].recomputeMeans(features);
// System.out.println(Arrays.toString(clusters[i].center));
}
// calculate stress
double stress = 0d;
double num = 0;
for (int i = 0; i < clusters.length; i++) {
stress += clusters[i].calculateStress(features);
num += clusters[i].members.size();
}
return stress;
}
public List<double[]> getMeans() {
ArrayList<double[]> r = new ArrayList<double[]>(clusters.length);
for (int i = 0; i < clusters.length; i++) {
r.add(clusters[i].center);
}
return r;
}
/**
* Cluster implementation used in this k-means implementation.
*/
class Cluster {
double[] center;
HashSet<Integer> members;
public Cluster(double[] center) {
this.center = center;
members = new HashSet<Integer>();
}
public double getDistance(double[] feature) {
return MetricsUtils.distL2(center, feature);
}
public void clearMembers() {
members.clear();
}
public void addMember(int id) {
members.add(id);
}
public void recomputeMeans(List<double[]> features) {
if (members.size() > 0) {
Arrays.fill(center, 0d);
for (Iterator<Integer> iterator = members.iterator(); iterator.hasNext(); ) {
int member = iterator.next();
double[] feature = features.get(member);
for (int i = 0; i < feature.length; i++) {
// assert (feature[i] < 256);
center[i] += feature[i];
}
}
for (int i = 0; i < center.length; i++) {
center[i] = center[i] / ((double) members.size());
}
}
}
public double calculateStress(List<double[]> features) {
double result = 0d;
for (Iterator<Integer> iterator = members.iterator(); iterator.hasNext(); ) {
int member = iterator.next();
double[] feature = features.get(member);
result += MetricsUtils.distL2(center, feature);
}
return result;
}
}
}