Examples of org.encog.engine.network.activation.ActivationFunction

Package org.encog.engine.network.activation

Examples of org.encog.engine.network.activation.ActivationFunction

org.encog.engine.network.activation.ActivationFunction
This interface allows various activation functions to be used with the neural network. Activation functions are applied to the output from each layer of a neural network. Activation functions scale the output into the desired range. Methods are provided both to process the activation function, as well as the derivative of the function. Some training algorithms, particularly back propagation, require that it be possible to take the derivative of the activation function. Not all activation functions support derivatives. If you implement an activation function that is not derivable then an exception should be thrown inside of the derivativeFunction method implementation. Non-derivable activation functions are perfectly valid, they simply cannot be used with every training algorithm.

   * @param network
   *            The network to compile for.
   */
  public void compile(final FlatNetwork network) {


    final ActivationFunction activation = network.getActivationFunctions()[0];
    final StringBuilder source = new StringBuilder();


    source.append("#define ACTIVATION(x,slope)");
    source.append(activation.getOpenCLExpression(false));
    source.append("\r\n");


    source.append(ResourceLoader.loadString(getSourceName()));
    setCLSource(source.toString());

View Full Code Here

   *            The network to compile for.
   */
  public void compile(final Map<String, String> options,
      final OpenCLTrainingProfile profile, final FlatNetwork network) {


    final ActivationFunction activation = network.getActivationFunctions()[0];
    final StringBuilder source = new StringBuilder();


    source.append("#define ACTIVATION(x,slope)");
    source.append(activation.getOpenCLExpression(false));
    source.append("\r\n");


    source.append("#define DERIVATIVE(x,slope)");
    source.append(activation.getOpenCLExpression(true));
    source.append("\r\n");


    source.append(ResourceLoader.loadString(getSourceName()));
    setCLSource(source.toString());

View Full Code Here

      value = params.get(name);
      if (value == null) {
        throw new PersistError("Missing property: " + name);
      }


      ActivationFunction af = null;
      final String[] cols = value.split("\\|");


      final String afName = "org.encog.engine.network.activation." + cols[0];
      try {
        final Class<?> clazz = Class.forName(afName);
        af = (ActivationFunction) clazz.newInstance();
      } catch (final ClassNotFoundException e) {
        throw new PersistError(e);
      } catch (final InstantiationException e) {
        throw new PersistError(e);
      } catch (final IllegalAccessException e) {
        throw new PersistError(e);
      }


      for (int i = 0; i < af.getParamNames().length; i++) {
        af.setParam(i, CSVFormat.EG_FORMAT.parse(cols[i + 1]));
      }


      return af;


    } catch (final Exception ex) {

View Full Code Here


  }


  public void changeEquation() {
    boolean der = this.derivative.isSelected();
    ActivationFunction newActivation = null;
    
    switch (this.select.getSelectedIndex()) {
    case 0:
      newActivation = new ActivationBiPolar();
      break;
    case 1:
      newActivation = new ActivationCompetitive();
      break;
    case 2:
      newActivation = new ActivationGaussian(0, 1, 1);
      break;
    case 3:
      newActivation = new ActivationLinear();
      break;
    case 4:
      newActivation = new ActivationLOG();
      break;
    case 5:
      newActivation = new ActivationSigmoid();
      break;
    case 6:
      newActivation = new ActivationSIN();
      break;
    case 7:
      newActivation = new ActivationSoftMax();
      break;
    case 8:
      newActivation = new ActivationStep();
      break;            
    case 9:
      newActivation = new ActivationTANH();
      break;
    case 10:
      newActivation = new ActivationRamp();
      break;      


    }
    
    if( this.activation.getClass() != newActivation.getClass() )
    {
      this.activation = newActivation;
    }
    
    this.equation.setupEquation(newActivation,!der);

View Full Code Here


  public ActivationFunction create(String fn) {
    
    for (EncogPluginBase plugin : Encog.getInstance().getPlugins()) {
      if (plugin instanceof EncogPluginService1) {
        ActivationFunction result = ((EncogPluginService1) plugin).createActivationFunction(fn);
        if (result != null) {
          return result;
        }
      }
    }

View Full Code Here

    }
    
    
    final BasicNetwork result = new BasicNetwork();
    final List<String> layers = ArchitectureParse.parseLayers(architecture);
    ActivationFunction af = new ActivationLinear();


    int questionPhase = 0;
    for (final String layerStr : layers) {
      int defaultCount;
      // determine default
      if (questionPhase == 0) {
        defaultCount = input;
      } else {
        defaultCount = output;
      }


      final ArchitectureLayer layer = ArchitectureParse.parseLayer(
          layerStr, defaultCount);
      final boolean bias = layer.isBias();


      String part = layer.getName();
      if (part != null) {
        part = part.trim();
      } else {
        part = "";
      }
      
      ActivationFunction lookup = this.factory.create(part);
      
      if (lookup!=null) {
        af = lookup;
      } else {
        if (layer.isUsedDefault()) {

View Full Code Here

  private void restructureFeedforward() {
    CreateFeedforward dialog = new CreateFeedforward(EncogWorkBench
        .getInstance().getMainWindow());
    BasicNetwork network = (BasicNetwork)method;


    ActivationFunction oldActivationOutput = network.getActivation(network
        .getLayerCount() - 1);
    dialog.setActivationFunctionOutput(oldActivationOutput);
    dialog.getInputCount().setValue(network.getInputCount());
    dialog.getOutputCount().setValue(network.getOutputCount());
    int hiddenLayerCount = network.getLayerCount() - 2;


    ActivationFunction oldActivationHidden = new ActivationTANH();
    for (int i = 0; i < hiddenLayerCount; i++) {
      int num = network.getLayerNeuronCount(i + 1);
      String str = "Hidden Layer " + (i + 1) + ": " + num + " neurons";
      dialog.getHidden().getModel().addElement(str);
    }

View Full Code Here

    final int toLayerIndex = this.layerIndex[currentLevel];
    final int fromLayerSize = this.layerCounts[currentLevel + 1];
    final int toLayerSize = this.layerFeedCounts[currentLevel];


    final int index = this.weightIndex[currentLevel];
    final ActivationFunction activation = this.flat
        .getActivationFunctions()[currentLevel + 1];


    // handle weights
    int yi = fromLayerIndex;
    for (int y = 0; y < fromLayerSize; y++) {
      final double output = this.layerOutput[yi];
      double sum = 0;
      int xi = toLayerIndex;
      int wi = index + y;
      for (int x = 0; x < toLayerSize; x++) {
        derivative[wi] += output * this.layerDelta[xi];
        sum += this.weights[wi] * this.layerDelta[xi];
        wi += fromLayerSize;
        xi++;
      }


      this.layerDelta[yi] = sum
          * (activation.derivativeFunction(this.layerSums[yi],this.layerOutput[yi]));
      yi++;
    }
  }

View Full Code Here

    for (int i = 0; i < activationFunctions.length; i++) {
      if (i > 0) {
        result.append(',');
      }


      final ActivationFunction af = activationFunctions[i];
      if (af instanceof ActivationSigmoid) {
        result.append("ENCOG.ActivationSigmoid.create()");
      } else if (af instanceof ActivationTANH) {
        result.append("ENCOG.ActivationTANH.create()");
      } else if (af instanceof ActivationLinear) {
        result.append("ENCOG.ActivationLinear.create()");
      } else if (af instanceof ActivationElliott) {
        result.append("ENCOG.ActivationElliott.create()");
      } else if (af instanceof ActivationElliott) {
        result.append("ENCOG.ActivationElliott.create()");
      } else {
        throw new AnalystCodeGenerationError(
            "Unsupported activatoin function for code generation: "
                + af.getClass().getSimpleName());
      }


    }
    result.append(']');
    return result.toString();

View Full Code Here

   * @return The array of flat activations.
   */
  public int[] createActivations(final FlatNetwork flat) {
    final int[] result = new int[flat.getActivationFunctions().length];
    for (int i = 0; i < flat.getActivationFunctions().length; i++) {
      final ActivationFunction af = flat.getActivationFunctions()[i];


      if (af instanceof ActivationLinear) {
        result[i] = 0;
      } else if (af instanceof ActivationTANH) {
        result[i] = 1;

View Full Code Here

0 1 2 3

TOP

Related Classes of org.encog.engine.network.activation.ActivationFunction

org.encog.app.generate.generators.AbstractTemplateGenerator

org.encog.app.generate.generators.js.GenerateEncogJavaScript

org.encog.engine.network.flat.FlatNetwork

org.encog.engine.network.train.gradient.GradientWorkerCPU

org.encog.engine.opencl.kernels.KernelNetworkCalc

org.encog.engine.opencl.kernels.KernelNetworkTrain

org.encog.mathutil.matrices.hessian.ChainRuleWorker

org.encog.mathutil.randomize.NguyenWidrowRandomizer

org.encog.ml.factory.method.FeedforwardFactory

org.encog.ml.factory.method.NEATFactory

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.