org.joone.engine
Class TanhLayer

java.lang.Object
  extended by org.joone.engine.Layer
      extended by org.joone.engine.SimpleLayer
          extended by org.joone.engine.TanhLayer
All Implemented Interfaces:
java.io.Serializable, java.lang.Runnable, Learnable, LearnableLayer, NeuralLayer, Inspectable

public class TanhLayer
extends SimpleLayer
implements LearnableLayer

Layer that applies the tangent hyperbolic transfer function to its input patterns

See Also:
Serialized Form

Field Summary
 
Fields inherited from class org.joone.engine.Layer
bias, gradientInps, gradientOuts, inps, inputPatternListeners, learnable, learning, m_batch, monitor, myLearner, outputPatternListeners, outs, running, step, STOP_FLAG
 
Constructor Summary
TanhLayer()
          default constructor
TanhLayer(java.lang.String name)
           
 
Method Summary
 void backward(double[] pattern)
          Reverse transfer function of the component.
 void forward(double[] pattern)
          Transfer function to recall a result on a trained net
 double getDefaultState()
          Return the default state of a node in this layer, such as 0 for a tanh or 0.5 for a sigmoid layer
 double getDerivative(int i)
          Similar to the backward message and used by RTRL
 double getFlatSpotConstant()
          Gets the flat spot constant.
 Learner getLearner()
          Deprecated. - Used only for backward compatibility
 double getMaximumState()
          Return maximum value of a node in this layer
 double getMinimumState()
          Return minimum value of a node in this layer
 void setFlatSpotConstant(double aConstant)
          Sets the constant to overcome the flat spot problem.
 
Methods inherited from class org.joone.engine.SimpleLayer
getLearningRate, getLrate, getMomentum, setDimensions, setLrate, setMomentum, setMonitor
 
Methods inherited from class org.joone.engine.Layer
addInputSynapse, addNoise, addOutputSynapse, adjustSizeToFwdPattern, adjustSizeToRevPattern, check, checkInputEnabled, checkInputs, checkOutputs, copyInto, finalize, fireFwdGet, fireFwdPut, fireRevGet, fireRevPut, fwdRun, getAllInputs, getAllOutputs, getBias, getDimension, getLastGradientInps, getLastGradientOuts, getLastInputs, getLastOutputs, getLayerName, getMonitor, getRows, getThreadMonitor, hasStepCounter, init, initLearner, InspectableTitle, Inspections, isInputLayer, isOutputLayer, isRunning, join, randomize, randomizeBias, randomizeWeights, removeAllInputs, removeAllOutputs, removeInputSynapse, removeListener, removeOutputSynapse, resetInputListeners, revRun, run, setAllInputs, setAllOutputs, setBias, setConnDimensions, setInputDimension, setInputSynapses, setLastInputs, setLastOutputs, setLayerName, setOutputDimension, setOutputSynapses, setRows, start, stop, sumBackInput, sumInput, toString
 
Methods inherited from class java.lang.Object
clone, equals, getClass, hashCode, notify, notifyAll, wait, wait, wait
 
Methods inherited from interface org.joone.engine.Learnable
getMonitor, initLearner
 
Methods inherited from interface org.joone.engine.NeuralLayer
addInputSynapse, addNoise, addOutputSynapse, check, copyInto, getAllInputs, getAllOutputs, getBias, getLayerName, getMonitor, getRows, isRunning, removeAllInputs, removeAllOutputs, removeInputSynapse, removeOutputSynapse, setAllInputs, setAllOutputs, setBias, setLayerName, setMonitor, setRows, start
 

Constructor Detail

TanhLayer

public TanhLayer()
default constructor


TanhLayer

public TanhLayer(java.lang.String name)
Method Detail

backward

public void backward(double[] pattern)
Description copied from class: Layer
Reverse transfer function of the component.

Overrides:
backward in class SimpleLayer
Parameters:
pattern - input pattern on which to apply the transfer function
See Also:
(double[])

getDerivative

public double getDerivative(int i)
Similar to the backward message and used by RTRL

Specified by:
getDerivative in class Layer

forward

public void forward(double[] pattern)
Description copied from class: Layer
Transfer function to recall a result on a trained net

Specified by:
forward in class Layer
Parameters:
pattern - input pattern to which to apply the rtransfer function
See Also:
(double[])

getLearner

public Learner getLearner()
Deprecated. - Used only for backward compatibility

Description copied from class: Layer
Returns the appropriate Learner object for this class depending on the Monitor.learningMode property value

Specified by:
getLearner in interface Learnable
Overrides:
getLearner in class Layer
Returns:
the Learner object if applicable, otherwise null
See Also:
Learnable.getLearner()

setFlatSpotConstant

public void setFlatSpotConstant(double aConstant)
Sets the constant to overcome the flat spot problem. This problem is described in: S.E. Fahlman, "An emperical study of learning speed in backpropagation with good scaling properties," Dept. Comput. Sci. Carnegie Mellon Univ., Pittsburgh, PA, Tech. Rep., CMU-CS-88-162, 1988. Setting this constant to 0 (default value), the derivative of the sigmoid function is unchanged (normal function). An good value for this constant might be 0.1.

Parameters:
aConstant -

getFlatSpotConstant

public double getFlatSpotConstant()
Gets the flat spot constant.

Returns:
the flat spot constant.

getDefaultState

public double getDefaultState()
Description copied from class: Layer
Return the default state of a node in this layer, such as 0 for a tanh or 0.5 for a sigmoid layer

Specified by:
getDefaultState in class Layer

getMinimumState

public double getMinimumState()
Description copied from class: Layer
Return minimum value of a node in this layer

Specified by:
getMinimumState in class Layer

getMaximumState

public double getMaximumState()
Description copied from class: Layer
Return maximum value of a node in this layer

Specified by:
getMaximumState in class Layer


Submit Feedback to pmarrone@users.sourceforge.net