halfdan/eva2
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Remove unused KMeans clustering (half broken implementation anyway)

Browse Source
This commit is contained in:
Fabian Becker 2015-12-27 15:04:38 +01:00
parent a4f10c8b71
commit 74e5dbd532
2 changed files with 0 additions and 187 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

33
src/main/java/eva2/tools/Cluster.java
View File

@ -1,33 +0,0 @@
package eva2.tools;
/**
* This class represents a cluster object in the R^N.
*/
public class Cluster {
/**
* Number of samples in the cluster.
*/
public int samplesInCluster;
/**
* Center of the cluster.
*/
public double[] center;
/**
* nearest sample (double[]) to the center of the cluster.
*/
public double[] nearestSample;
/**
* This class represents a cluster of
* sample points.
*
* @param center center
* @param SamplesInCluster Number of samples in cluster
* @param nearestSample Nearest sample to cluster center.
*/
public Cluster(double[] center, int SamplesInCluster, double[] nearestSample) {
samplesInCluster = SamplesInCluster;
this.center = center;
this.nearestSample = nearestSample;
}
}

154
src/main/java/eva2/tools/KMEANSJAVA.java
View File

@ -1,154 +0,0 @@
package eva2.tools;
import eva2.tools.math.RNG;
import java.util.Comparator;
/**
*
*/
public class KMEANSJAVA {
protected double[][] c;
protected int[] indices;
/**
*
*/
public double[][] getC() {
return c;
}
/**
*
*/
public int[] getIDX() {
return indices;
}
/**
*
*/
private double dist(double[] x1, double[] x2) {
double ret = 0;
for (int i = 0; i < x1.length; i++) {
ret += (x1[i] - x2[i]) * (x1[i] - x2[i]);
}
return Math.sqrt(ret);
}
/**
*
*/
public KMEANSJAVA(double[][] samples, int K, int iterations) {
if (K > samples.length) {
K = samples.length;
}
int counter = 0;
c = new double[K][];
for (int i = 0; i < K; i++) {
c[i] = samples[i].clone();
}
indices = new int[samples.length];
while (counter++ < iterations) {
// determine indices start
for (int i = 0; i < indices.length; i++) {
int index_nc = 0; // index of nearest cluster
double mindist = 999999999;
for (int j = 0; j < c.length; j++) {
if (mindist > dist(samples[i], c[j])) {
mindist = dist(samples[i], c[j]);
index_nc = j;
}
}
indices[i] = index_nc;
}
// determine indices end !
// determine the new centers
for (int indexofc = 0; indexofc < c.length; indexofc++) {
double[] newcenter = new double[samples[0].length];
int treffer = 0;
for (int indice : indices) { //System.out.println("j="+j);
if (indice == indexofc) {
treffer++;
for (int d = 0; d < newcenter.length; d++) {
newcenter[d] += c[indice][d];
//newcenter[d] = newcenter[d] + samples[j][d];
}
}
}
for (int d = 0; d < newcenter.length; d++) {
newcenter[d] /= treffer;
}
c[indexofc] = newcenter;
}
// determine the new centers
}
}
/**
* Just a test function.
*/
public static void main(String[] args) {
int k = 3;
int samples = 10;
int d = 2;
double[][] test = new double[samples][d];
for (int i = 0; i < samples; i++) {
for (int j = 0; j < d; j++) {
test[i][j] = RNG.randomDouble(0, 10);
}
}
KMEANSJAVA app = new KMEANSJAVA(test, k, 5);
double[][] c = app.getC();
int[] idx = app.getIDX();
System.out.println("c");
for (int i = 0; i < c.length; i++) {
for (int j = 0; j < c[i].length; j++) {
System.out.print(c[i][j] + " ");
}
System.out.println("");
}
System.out.println("test");
for (double[] aTest : test) {
for (int j = 0; j < aTest.length; j++) {
System.out.print(aTest[j] + " ");
}
System.out.println("");
}
}
}
/**
*
*/
class ClusterComp implements Comparator {
/**
*
*/
public ClusterComp() {
}
/**
*
*/
@Override
public int compare(Object p1, Object p2) {
int x1 = ((Cluster) p1).samplesInCluster;
int x2 = ((Cluster) p2).samplesInCluster;
if (x1 > x2) {
return -1;
}
if (x1 <= x2) {
return 1;
}
return 0;
}
/**
*
*/
@Override
public boolean equals(Object x) {
return false;
}
}