weka.classifiers.neural.singlelayerperceptron.algorithm.SLPAlgorithmAncestor Maven / Gradle / Ivy

Go to download
Show more of this group Show more artifacts with this name
Show all versions of wekaclassalgos Show documentation
Fork of the following defunct sourceforge.net project: https://sourceforge.net/projects/wekaclassalgos/
There is a newer version: 2023.2.8
/*
 *   This program is free software: you can redistribute it and/or modify
 *   it under the terms of the GNU General Public License as published by
 *   the Free Software Foundation, either version 3 of the License, or
 *   (at your option) any later version.
 *
 *   This program is distributed in the hope that it will be useful,
 *   but WITHOUT ANY WARRANTY; without even the implied warranty of
 *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *   GNU General Public License for more details.
 *
 *   You should have received a copy of the GNU General Public License
 *   along with this program.  If not, see .
 */

package weka.classifiers.neural.singlelayerperceptron.algorithm;

import weka.classifiers.neural.common.CommonNeuralAlgorithmAncestor;
import weka.classifiers.neural.common.RandomWrapper;
import weka.classifiers.neural.common.SimpleNeuron;
import weka.classifiers.neural.common.initialisation.Initialisation;
import weka.classifiers.neural.common.learning.LearningRateKernel;
import weka.classifiers.neural.common.transfer.TransferFunction;
import weka.core.Instance;
import weka.core.Instances;

/**
 * Title: Weka Neural Implementation
 * Description: ...
 * Copyright: Copyright (c) 2003
 * Company: N/A
 *
 * @author Jason Brownlee
 * @version 1.0
 */

public abstract class SLPAlgorithmAncestor extends CommonNeuralAlgorithmAncestor {

  // neurons which make up this model
  protected final SimpleNeuron[] neurons;

  // learning rate function
  protected final LearningRateKernel learningRateFunction;


  public SLPAlgorithmAncestor(TransferFunction aTransfer,
			      double aBiasInput,
			      RandomWrapper aRand,
			      LearningRateKernel aKernel,
			      Instances trainingInstances) {
    super(aTransfer, aRand);

    learningRateFunction = aKernel;

    // determine the number of neurons required
    if (trainingInstances.classAttribute().isNumeric()) {
      neurons = new SimpleNeuron[1];
    }
    // must be numeric
    else {
      neurons = new SimpleNeuron[trainingInstances.numClasses()];
    }

    // prepare the network structure
    prepareNetworkStructure(trainingInstances, aBiasInput);
  }


  public int getNumOutputNeurons() {
    if (neurons == null) {
      return 0;
    }

    return neurons.length;
  }


  public double[] getAllWeights() {
    if (neurons == null) {
      return null;
    }

    int totalWeights = neurons.length * neurons[0].getWeights().length;
    double[] weights = new double[totalWeights];
    int offset = 0;

    for (int i = 0; i < neurons.length; i++) {
      double[] tmpWeights = neurons[i].getWeights();

      for (int k = 0; k < tmpWeights.length; k++, offset++) {
	weights[offset] = tmpWeights[k];
      }
    }

    return weights;
  }


  protected abstract void calculateWeightErrors(Instance instance, SimpleNeuron neuron, double expected, double aLearningRate);


  public double getLearningRate(int aEpochNumber) {
    return learningRateFunction.currentLearningRate(aEpochNumber);
  }

  public void startingEpoch() {
  }

  public void finishedEpoch(Instances instances, double aLearningRate) {
  }


  public String getModelInformation() {
    StringBuffer buffer = new StringBuffer();

    buffer.append("Initial Learing Rate     : " + learningRateFunction.getInitialLearningRate() + "\n");
    buffer.append("Bias Input Value         : " + neurons[0].getBiasInputValue() + "\n");
    buffer.append("Output Layer Neurons     : " + neurons.length + "\n");

    return buffer.toString();
  }


  public void updateModel(Instance inputs, double aLearningRate) {
    // prepare an expected output vector
    double[] expected = prepareExpectedOutputVector(inputs);

    // calculate weight changes for each neuron
    for (int i = 0; i < neurons.length; i++) {
      // calculate weight changes
      calculateWeightErrors(inputs, neurons[i], expected[i], aLearningRate);
    }
  }

  protected void prepareNetworkStructure(Instances instances, double aBiasInput) {
    // determine the number of attribtes
    int numAttributes = instances.numAttributes() - 1;

    // construct the required number of neurons
    for (int i = 0; i < neurons.length; i++) {
      neurons[i] = new SimpleNeuron(numAttributes, aBiasInput);

      // initialise weights to between -0.5 and +0.5
      Initialisation.initialiseVectorToRandomWithSign(neurons[i].getWeights(), 0.5, 0.0, rand);
    }
  }


  public double[] getNetworkOutputs(Instance instance) {
    double[] distribution = new double[neurons.length];

    for (int i = 0; i < distribution.length; i++) {
      double activation = activate(neurons[i], instance);
      distribution[i] = transfer(activation);
    }

    return distribution;
  }

}