public NeuralNetwork(int input, int hidden, int output) {
    this.layers = new int[] {input, hidden, output};
    df = new DecimalFormat("#0.000#");
    ef = new DecimalFormat("#");

    /** Create all neurons and connections Connections are created in the neuron class */
    for (int i = 0; i < layers.length; i++) {
      if (i == 0) { // input layer
        for (int j = 0; j < layers[i]; j++) {
          Neuron neuron = new Neuron();
          inputLayer.add(neuron);
        }
      } else if (i == 1) { // hidden layer
        for (int j = 0; j < layers[i]; j++) {
          Neuron neuron = new Neuron();
          neuron.addInConnectionsS(inputLayer);
          neuron.addBiasConnection(bias);
          hiddenLayer.add(neuron);
        }
      } else if (i == 2) { // output layer
        for (int j = 0; j < layers[i]; j++) {
          Neuron neuron = new Neuron();
          neuron.addInConnectionsS(hiddenLayer);
          neuron.addBiasConnection(bias);
          outputLayer.add(neuron);
        }
      } else {
        System.out.println("!Error NeuralNetwork init");
      }
    }

    // initialize random weights
    for (Neuron neuron : hiddenLayer) {
      ArrayList<Connection> connections = neuron.getAllInConnections();
      for (Connection conn : connections) {
        double newWeight = getRandom();
        conn.setWeight(newWeight);
      }
    }
    for (Neuron neuron : outputLayer) {
      ArrayList<Connection> connections = neuron.getAllInConnections();
      for (Connection conn : connections) {
        double newWeight = getRandom();
        conn.setWeight(newWeight);
      }
    }

    // reset id counters
    Neuron.counter = 0;
    Connection.counter = 0;

    if (isTrained) {
      trainedWeights();
      updateAllWeights();
    }
  }
  /**
   * all output propagate back
   *
   * @param expectedOutput first calculate the partial derivative of the error with respect to each
   *     of the weight leading into the output neurons bias is also updated here
   */
  public void applyBackpropagation(double expectedOutput[]) {

    // error check, normalize value ]0;1[
    for (int i = 0; i < expectedOutput.length; i++) {
      double d = expectedOutput[i];
      if (d < 0 || d > 1) {
        if (d < 0) expectedOutput[i] = 0 + epsilon;
        else expectedOutput[i] = 1 - epsilon;
      }
    }

    int i = 0;
    for (Neuron n : outputLayer) {
      ArrayList<Connection> connections = n.getAllInConnections();
      for (Connection con : connections) {
        double ak = n.getOutput();
        double ai = con.leftNeuron.getOutput();
        double desiredOutput = expectedOutput[i];

        double partialDerivative = -ak * (1 - ak) * ai * (desiredOutput - ak);
        double deltaWeight = -learningRate * partialDerivative;
        double newWeight = con.getWeight() + deltaWeight;
        con.setDeltaWeight(deltaWeight);
        con.setWeight(newWeight + momentum * con.getPrevDeltaWeight());
      }
      i++;
    }

    // update weights for the hidden layer
    for (Neuron n : hiddenLayer) {
      ArrayList<Connection> connections = n.getAllInConnections();
      for (Connection con : connections) {
        double aj = n.getOutput();
        double ai = con.leftNeuron.getOutput();
        double sumKoutputs = 0;
        int j = 0;
        for (Neuron out_neu : outputLayer) {
          double wjk = out_neu.getConnection(n.id).getWeight();
          double desiredOutput = (double) expectedOutput[j];
          double ak = out_neu.getOutput();
          j++;
          sumKoutputs = sumKoutputs + (-(desiredOutput - ak) * ak * (1 - ak) * wjk);
        }

        double partialDerivative = aj * (1 - aj) * ai * sumKoutputs;
        double deltaWeight = -learningRate * partialDerivative;
        double newWeight = con.getWeight() + deltaWeight;
        con.setDeltaWeight(deltaWeight);
        con.setWeight(newWeight + momentum * con.getPrevDeltaWeight());
      }
    }
  }
Beispiel #3
0
 private void mutateWeights() {
   for (Connection connection : this.connections) {
     if (this.randomGenerator.nextDouble() > 0.8) {
       // 80% chance of mutation
       continue;
     }
     if (this.randomGenerator.nextDouble() > 0.9) {
       // 10% chance of random reset
       connection.setWeight(NNGenome.scale(this.randomGenerator.nextDouble(), 0, 1, -1, 1));
     } else { // 90% chance of perturbation
       double perturbation = NNGenome.scale(this.randomGenerator.nextDouble(), 0, 1, -0.01, 0.01);
       connection.setWeight(connection.getWeight() + perturbation);
     }
   }
 }
 /** Take from hash table and put into all weights */
 public void updateAllWeights() {
   // update weights for the output layer
   for (Neuron n : outputLayer) {
     ArrayList<Connection> connections = n.getAllInConnections();
     for (Connection con : connections) {
       String key = weightKey(n.id, con.id);
       double newWeight = weightUpdate.get(key);
       con.setWeight(newWeight);
     }
   }
   // update weights for the hidden layer
   for (Neuron n : hiddenLayer) {
     ArrayList<Connection> connections = n.getAllInConnections();
     for (Connection con : connections) {
       String key = weightKey(n.id, con.id);
       double newWeight = weightUpdate.get(key);
       con.setWeight(newWeight);
     }
   }
 }