/*
* Encog(tm) Core v3.3 - Java Version
* http://www.heatonresearch.com/encog/
* https://github.com/encog/encog-java-core
* Copyright 2008-2014 Heaton Research, Inc.
*
* 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.
*
* For more information on Heaton Research copyrights, licenses
* and trademarks visit:
* http://www.heatonresearch.com/copyright
*/
package org.encog.neural.neat.training.opp;
import java.util.Random;
import org.encog.engine.network.activation.ActivationFunction;
import org.encog.mathutil.randomize.RangeRandomizer;
import org.encog.ml.ea.genome.Genome;
import org.encog.neural.neat.NEATNeuronType;
import org.encog.neural.neat.NEATPopulation;
import org.encog.neural.neat.training.NEATGenome;
import org.encog.neural.neat.training.NEATInnovation;
import org.encog.neural.neat.training.NEATLinkGene;
import org.encog.neural.neat.training.NEATNeuronGene;
/**
* Mutate a genome by adding a new node. To do this a random link is chosen. The
* a neuron is created to split this link. This removes one link and adds two
* new links. The weights on the new link are created to minimize changes to the
* values produced by the neuron.
*
* -----------------------------------------------------------------------------
* http://www.cs.ucf.edu/~kstanley/ Encog's NEAT implementation was drawn from
* the following three Journal Articles. For more complete BibTeX sources, see
* NEATNetwork.java.
*
* Evolving Neural Networks Through Augmenting Topologies
*
* Generating Large-Scale Neural Networks Through Discovering Geometric
* Regularities
*
* Automatic feature selection in neuroevolution
*/
public class NEATMutateAddNode extends NEATMutation {
/**
* {@inheritDoc}
*/
@Override
public void performOperation(final Random rnd, final Genome[] parents,
final int parentIndex, final Genome[] offspring,
final int offspringIndex) {
final NEATGenome target = obtainGenome(parents, parentIndex, offspring,
offspringIndex);
int countTrysToFindOldLink = getOwner().getMaxTries();
final NEATPopulation pop = ((NEATPopulation) target.getPopulation());
// the link to split
NEATLinkGene splitLink = null;
final int sizeBias = ((NEATGenome)parents[0]).getInputCount()
+ ((NEATGenome)parents[0]).getOutputCount() + 10;
// if there are not at least
int upperLimit;
if (target.getLinksChromosome().size() < sizeBias) {
upperLimit = target.getNumGenes() - 1
- (int) Math.sqrt(target.getNumGenes());
} else {
upperLimit = target.getNumGenes() - 1;
}
while ((countTrysToFindOldLink--) > 0) {
// choose a link, use the square root to prefer the older links
final int i = RangeRandomizer.randomInt(0, upperLimit);
final NEATLinkGene link = target.getLinksChromosome().get(i);
// get the from neuron
final long fromNeuron = link.getFromNeuronID();
if ((link.isEnabled())
&& (target.getNeuronsChromosome()
.get(getElementPos(target, fromNeuron))
.getNeuronType() != NEATNeuronType.Bias)) {
splitLink = link;
break;
}
}
if (splitLink == null) {
return;
}
splitLink.setEnabled(false);
final long from = splitLink.getFromNeuronID();
final long to = splitLink.getToNeuronID();
final NEATInnovation innovation = ((NEATPopulation)getOwner().getPopulation()).getInnovations()
.findInnovationSplit(from, to);
// add the splitting neuron
final ActivationFunction af = ((NEATPopulation)getOwner().getPopulation())
.getActivationFunctions().pick(new Random());
target.getNeuronsChromosome().add(
new NEATNeuronGene(NEATNeuronType.Hidden, af, innovation
.getNeuronID(), innovation.getInnovationID()));
// add the other two sides of the link
createLink(target, from, innovation.getNeuronID(),
splitLink.getWeight());
createLink(target, innovation.getNeuronID(), to, pop.getWeightRange());
target.sortGenes();
}
}