/**
* Copyright 2010 Neuroph Project http://neuroph.sourceforge.net
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.neuroph.nnet.learning;
import java.util.Iterator;
import org.neuroph.core.Connection;
import org.neuroph.core.Layer;
import org.neuroph.core.NeuralNetwork;
import org.neuroph.core.Neuron;
import org.neuroph.core.learning.LearningRule;
import org.neuroph.core.learning.TrainingElement;
import org.neuroph.core.learning.TrainingSet;
import org.neuroph.nnet.Kohonen;
/**
* Learning algorithm for Kohonen network.
*
* @author Zoran Sevarac <sevarac@gmail.com>
*/
public class KohonenLearning extends LearningRule {
/**
* The class fingerprint that is set to indicate serialization
* compatibility with a previous version of the class.
*/
private static final long serialVersionUID = 1L;
double learningRate = 0.9d;
int[] iterations = { 100, 0 };
double decStep[] = new double[2];
int mapSize = 0;
int[] nR = { 1, 1 }; // neighborhood radius
int currentIteration;
public KohonenLearning() {
super();
}
public void learn(TrainingSet trainingSet) {
for (int phase = 0; phase < 2; phase++) {
for (int k = 0; k < iterations[phase]; k++) {
Iterator<TrainingElement> e = trainingSet.iterator();
while (e.hasNext() && !isStopped()) {
TrainingElement tE = e.next();
learnPattern(tE, nR[phase]);
} // while
currentIteration = k;
this.notifyChange();
if (isStopped()) return;
} // for k
learningRate = learningRate * 0.5;
} // for phase
}
private void learnPattern(TrainingElement tE, int neighborhood) {
neuralNetwork.setInput(tE.getInput());
neuralNetwork.calculate();
Neuron winner = getClosest();
if (winner.getOutput() == 0)
return; // ako je vec istrenirana jedna celija, izadji
Layer mapLayer = neuralNetwork.getLayerAt(1);
int winnerIdx = mapLayer.indexOf(winner);
adjustCellWeights(winner, 0);
int cellNum = mapLayer.getNeuronsCount();
for (int p = 0; p < cellNum; p++) {
if (p == winnerIdx)
continue;
if (isNeighbor(winnerIdx, p, neighborhood)) {
Neuron cell = mapLayer.getNeuronAt(p);
adjustCellWeights(cell, 1);
} // if
} // for
}
// get unit with closetst weight vector
private Neuron getClosest() {
Iterator<Neuron> i = this.neuralNetwork.getLayerAt(1)
.getNeuronsIterator();
Neuron winner = new Neuron();
double minOutput = 100;
while (i.hasNext()) {
Neuron n = i.next();
double out = n.getOutput();
if (out < minOutput) {
minOutput = out;
winner = n;
} // if
} // while
return winner;
}
private void adjustCellWeights(Neuron cell, int r) {
Iterator<Connection> i = cell.getInputsIterator();
while (i.hasNext()) {
Connection conn = i.next();
double dWeight = (learningRate / (r + 1))
* (conn.getInput() - conn.getWeight().getValue());
conn.getWeight().inc(dWeight);
}// while
}
private boolean isNeighbor(int i, int j, int n) {
// i - centralna celija
// n - velicina susedstva
// j - celija za proveru
n = 1;
int d = mapSize;
// if (j<(i-n*d-n)||(j>(i+n*d+n))) return false;
int rt = n; // broj celija ka gore
while ((i - rt * d) < 0)
rt--;
int rb = n; // broj celija ka dole
while ((i + rb * d) > (d * d - 1))
rb--;
for (int g = -rt; g <= rb; g++) {
int rl = n; // broj celija u levu stranu
int rl_mod = (i - rl) % d;
int i_mod = i % d;
while (rl_mod > i_mod) {
rl--;
rl_mod = (i - rl) % d;
}
int rd = n; // broj celija u desnu stranu
int rd_mod = (i + rd) % d;
while (rd_mod < i_mod) {
rd--;
rd_mod = (i + rd) % d;
}
if ((j >= (i + g * d - rl)) && (j <= (i + g * d + rd)))
return true;
// else if (j<(i+g*d-rl)) return false;
} // for
return false;
}
public double getLearningRate() {
return learningRate;
}
public void setLearningRate(double learningRate) {
this.learningRate = learningRate;
}
public void setIterations(int Iphase, int IIphase) {
this.iterations[0] = Iphase;
this.iterations[1] = IIphase;
}
public Integer getIteration() {
return new Integer(currentIteration);
}
public int getMapSize() {
return mapSize;
}
@Override
public void setNeuralNetwork(NeuralNetwork neuralNetwork) {
super.setNeuralNetwork(neuralNetwork);
int neuronsNum = neuralNetwork.getLayerAt(1).getNeuronsCount();
mapSize = (int) Math.sqrt(neuronsNum);
}
}