Package org.neuroph.nnet.learning

Source Code of org.neuroph.nnet.learning.MomentumBackpropagation

/**
* 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 org.neuroph.core.learning.TrainingData;
import org.neuroph.core.Connection;
import org.neuroph.core.NeuralNetwork;
import org.neuroph.core.Neuron;
import org.neuroph.core.Weight;


/**
* Backpropagation learning rule with momentum.
* @author Zoran Sevarac <sevarac@gmail.com>
*/
public class MomentumBackpropagation extends BackPropagation {
 
  /**
   * The class fingerprint that is set to indicate serialization
   * compatibility with a previous version of the class.
   */ 
  private static final long serialVersionUID = 1L;
 
  /**
   * Momentum factor
   */
  protected double momentum = 0.25d;

    
  /**
   * Creates new instance of MomentumBackpropagation learning
         */
  public MomentumBackpropagation() {
    super();
                this.setTrainingDataBufferSize(2); // batch weights sum and previous weight value
  }


  /**
   * This method implements weights update procedure for the single neuron
   * for the back propagation with momentum factor
   * @param neuron
   *            neuron to update weights
   */
  @Override
  protected void updateNeuronWeights(Neuron neuron) {
    for(Connection connection : neuron.getInputConnections() ) {
      double input = connection.getInput();
      if (input == 0) {
        continue;
      }

                        // get the error for specified neuron,
                        double neuronError = neuron.getError();

                        // tanh can be used to minimise the impact of big error values, which can cause network instability
                        // suggested at https://sourceforge.net/tracker/?func=detail&atid=1107579&aid=3130561&group_id=238532
                        // double neuronError = Math.tanh(neuron.getError());
     
      Weight weight = connection.getWeight();
     
      double currentWeighValue = weight.getValue();
      double previousWeightValue = weight.getTrainingData().get(TrainingData.PREVIOUS_WEIGHT);
      double deltaWeight = this.learningRate * neuronError * input +
        momentum * (currentWeighValue - previousWeightValue);
                        // save previous weight value
                        weight.getTrainingData().set(TrainingData.PREVIOUS_WEIGHT, currentWeighValue);
                        this.applyWeightChange(weight, deltaWeight);
    }
  }

  /**
   * Returns the momentum factor
   *
   * @return momentum factor
   */
  public double getMomentum() {
    return momentum;
  }

  /**
   * Sets the momentum factor
   *
   * @param momentum
   *            momentum factor
   */ 
  public void setMomentum(double momentum) {
    this.momentum = momentum;
  }

}
TOP

Related Classes of org.neuroph.nnet.learning.MomentumBackpropagation

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.