Package smile.base.mlp

Class InputLayer

java.lang.Object

smile.base.mlp.Layer

smile.base.mlp.InputLayer

All Implemented Interfaces:

Serializable, AutoCloseable

public class InputLayer
extends Layer

An input layer in the neural network.

See Also:

• Serialized Form

Field Summary

Fields inherited from class smile.base.mlp.Layer

bias, biasGradient, biasGradientMoment1, biasGradientMoment2, biasUpdate, dropout, mask, n, output, outputGradient, p, weight, weightGradient, weightGradientMoment1, weightGradientMoment2, weightUpdate

Constructor Summary

Constructors

Constructor	Description
InputLayer(int p)	Constructor.
InputLayer(int p, double dropout)	Constructor.

Method Summary

Modifier and Type	Method	Description
void	backpropagate(double[] lowerLayerGradient)	Propagates the errors back to a lower layer.
void	computeGradient(double[] x)	Computes the parameter gradient for a sample of (mini-)batch.
void	computeGradientUpdate(double[] x, double learningRate, double momentum, double decay)	Computes the parameter gradient and update the weights.
void	propagate(double[] x)	Propagates the signals from a lower layer to this layer.
String	toString()
void	transform(double[] x)	The activation or output function.
void	update(int m, double learningRate, double momentum, double decay, double rho, double epsilon)	Adjust network weights by back-propagation algorithm.

Methods inherited from class smile.base.mlp.Layer

backpopagateDropout, builder, close, getInputSize, getOutputSize, gradient, input, input, leaky, leaky, leaky, linear, linear, mle, mse, of, output, propagateDropout, rectifier, rectifier, sigmoid, sigmoid, tanh, tanh

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Constructor Details

InputLayer

public InputLayer(int p)

Constructor.

Parameters:

p - the number of input variables (not including bias value).

InputLayer

public InputLayer(int p,
 double dropout)

Constructor.

Parameters:

p - the number of input variables (not including bias value).

dropout - the dropout rate.

Method Details

toString

public String toString()

Overrides:

toString in class Object

propagate

public void propagate(double[] x)

Description copied from class: Layer

Propagates the signals from a lower layer to this layer.

Overrides:

propagate in class Layer

Parameters:

x - the lower layer signals.

backpropagate

public void backpropagate(double[] lowerLayerGradient)

Description copied from class: Layer

Propagates the errors back to a lower layer.

Specified by:

backpropagate in class Layer

Parameters:

lowerLayerGradient - the gradient vector of lower layer.

transform

public void transform(double[] x)

Description copied from class: Layer

The activation or output function.

Specified by:

transform in class Layer

Parameters:

x - the input and output values.

computeGradient

public void computeGradient(double[] x)

Description copied from class: Layer

Computes the parameter gradient for a sample of (mini-)batch.

Overrides:

computeGradient in class Layer

Parameters:

x - the input vector.

computeGradientUpdate

public void computeGradientUpdate(double[] x,
 double learningRate,
 double momentum,
 double decay)

Description copied from class: Layer

Computes the parameter gradient and update the weights.

Overrides:

computeGradientUpdate in class Layer

Parameters:

x - the input vector.

learningRate - the learning rate.

momentum - the momentum factor.

decay - weight decay factor.

update

public void update(int m,
 double learningRate,
 double momentum,
 double decay,
 double rho,
 double epsilon)

Description copied from class: Layer

Adjust network weights by back-propagation algorithm.

Overrides:

update in class Layer

Parameters:

m - the size of mini-batch.

learningRate - the learning rate.

momentum - the momentum factor.

decay - weight decay factor.

rho - RMSProp discounting factor for the history/coming gradient.

epsilon - a small constant for numerical stability.