Package org.encog.neural.hyperneat

Source Code of org.encog.neural.hyperneat.HyperNEATCODEC

/*
* 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.hyperneat;

import java.util.ArrayList;
import java.util.Collections;
import java.util.List;

import org.encog.engine.network.activation.ActivationFunction;
import org.encog.engine.network.activation.ActivationSteepenedSigmoid;
import org.encog.ml.MLMethod;
import org.encog.ml.data.MLData;
import org.encog.ml.data.basic.BasicMLData;
import org.encog.ml.ea.codec.GeneticCODEC;
import org.encog.ml.ea.genome.Genome;
import org.encog.ml.genetic.GeneticError;
import org.encog.neural.hyperneat.substrate.Substrate;
import org.encog.neural.hyperneat.substrate.SubstrateLink;
import org.encog.neural.hyperneat.substrate.SubstrateNode;
import org.encog.neural.neat.NEATCODEC;
import org.encog.neural.neat.NEATLink;
import org.encog.neural.neat.NEATNetwork;
import org.encog.neural.neat.NEATPopulation;

public class HyperNEATCODEC implements GeneticCODEC {

  private double minWeight = 0.2;
  private double maxWeight = 5.0;

  /**
   * {@inheritDoc}
   */
  @Override
  public MLMethod decode(final Genome genome) {
    final NEATPopulation pop = (NEATPopulation) genome.getPopulation();
    final Substrate substrate = pop.getSubstrate();
    return decode(pop, substrate, genome);
  }

  public MLMethod decode(final NEATPopulation pop, final Substrate substrate,
      final Genome genome) {
    // obtain the CPPN
    final NEATCODEC neatCodec = new NEATCODEC();
    final NEATNetwork cppn = (NEATNetwork) neatCodec.decode(genome);

    final List<NEATLink> linkList = new ArrayList<NEATLink>();

    final ActivationFunction[] afs = new ActivationFunction[substrate
        .getNodeCount()];

    final ActivationFunction af = new ActivationSteepenedSigmoid();
    // all activation functions are the same
    for (int i = 0; i < afs.length; i++) {
      afs[i] = af;
    }

    final double c = this.maxWeight / (1.0 - this.minWeight);
    final MLData input = new BasicMLData(cppn.getInputCount());

    // First create all of the non-bias links.
    for (final SubstrateLink link : substrate.getLinks()) {
      final SubstrateNode source = link.getSource();
      final SubstrateNode target = link.getTarget();

      int index = 0;
      for (final double d : source.getLocation()) {
        input.setData(index++, d);
      }
      for (final double d : target.getLocation()) {
        input.setData(index++, d);
      }
      final MLData output = cppn.compute(input);

      double weight = output.getData(0);
      if (Math.abs(weight) > this.minWeight) {
        weight = (Math.abs(weight) - this.minWeight) * c
            * Math.signum(weight);
        linkList.add(new NEATLink(source.getId(), target.getId(),
            weight));
      }
    }

    // now create biased links
    input.clear();
    final int d = substrate.getDimensions();
    final List<SubstrateNode> biasedNodes = substrate.getBiasedNodes();
    for (final SubstrateNode target : biasedNodes) {
      for (int i = 0; i < d; i++) {
        input.setData(d + i, target.getLocation()[i]);
      }

      final MLData output = cppn.compute(input);

      double biasWeight = output.getData(1);
      if (Math.abs(biasWeight) > this.minWeight) {
        biasWeight = (Math.abs(biasWeight) - this.minWeight) * c
            * Math.signum(biasWeight);
        linkList.add(new NEATLink(0, target.getId(), biasWeight));
      }
    }

    // check for invalid neural network
    if (linkList.size() == 0) {
      return null;
    }

    Collections.sort(linkList);

    final NEATNetwork network = new NEATNetwork(substrate.getInputCount(),
        substrate.getOutputCount(), linkList, afs);

    network.setActivationCycles(substrate.getActivationCycles());
    return network;

  }

  @Override
  public Genome encode(final MLMethod phenotype) {
    throw new GeneticError(
        "Encoding of a HyperNEAT network is not supported.");
  }

  /**
   * @return the maxWeight
   */
  public double getMaxWeight() {
    return this.maxWeight;
  }

  /**
   * @return the minWeight
   */
  public double getMinWeight() {
    return this.minWeight;
  }

  /**
   * @param maxWeight
   *            the maxWeight to set
   */
  public void setMaxWeight(final double maxWeight) {
    this.maxWeight = maxWeight;
  }

  /**
   * @param minWeight
   *            the minWeight to set
   */
  public void setMinWeight(final double minWeight) {
    this.minWeight = minWeight;
  }

}
TOP

Related Classes of org.encog.neural.hyperneat.HyperNEATCODEC

TOP
Copyright © 2018 www.massapi.com. All rights reserved.
All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.