Package me.yixqiao.jlearn.layers

Class Dense

java.lang.Object

me.yixqiao.jlearn.layers.Layer

me.yixqiao.jlearn.layers.Dense

All Implemented Interfaces:

java.io.Serializable

public class Dense
extends Layer

Basic fully-connected layer.

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field	Description
Matrix	biases	Biases.
Matrix	weights	Weights.

Constructor Summary

Constructors

Constructor	Description
Dense(int outSize, Activation activation)	Create a new dense layer.
Dense(int outSize, Activation activation, Initializer init)	Create a new dense layer.

Method Summary

Modifier and Type	Method	Description
Layer	cloneLayer()	Clone the layer, including weights and biases.
Layer	cloneSettings()	Clone the settings.
Matrix	forwardPropagate(Matrix x)	Forward propagate a batch of input.
Activation	getActivation()	Get the activation of the layer.
Matrix	getErrors(Matrix prevErrors)	Get backpropagated errors.
Matrix	getErrorsExpected(Matrix y)	Get errors from output layer.
int	getOutSize()	Get the output size of the layer.
void	initLayer(int inSize, Activation prevActivation)	Create the layer.
void	setOptimizers(Optimizer wOptimizer, Optimizer bOptimizer)	Set optimizers for training the layer.
java.lang.String	toString()	Get a string representation.
void	update(Matrix errors)	Update the layer after calculating errors.

Methods inherited from class java.lang.Object

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

Field Detail

weights

public Matrix weights

Weights.

biases

public Matrix biases

Biases.

Constructor Detail

Dense

public Dense(int outSize,
             Activation activation)

Create a new dense layer.

Parameters:

outSize - size of output

activation - activation to use

Dense

public Dense(int outSize,
             Activation activation,
             Initializer init)

Create a new dense layer.

Parameters:

outSize - size of output

activation - activation to use

init - initialization method

Method Detail

setOptimizers

public void setOptimizers(Optimizer wOptimizer,
                          Optimizer bOptimizer)

Description copied from class: Layer

Set optimizers for training the layer.

Specified by:

setOptimizers in class Layer

Parameters:

wOptimizer - optimizer for weights

bOptimizer - optimizer for biases

initLayer

public void initLayer(int inSize,
                      Activation prevActivation)

Description copied from class: Layer

Create the layer.

Specified by:

initLayer in class Layer

Parameters:

inSize - size of the previous layer

prevActivation - activation function or previous layer

getActivation

public Activation getActivation()

Description copied from class: Layer

Get the activation of the layer.

Specified by:

getActivation in class Layer

Returns:

the activation object

getOutSize

public int getOutSize()

Description copied from class: Layer

Get the output size of the layer.

Specified by:

getOutSize in class Layer

Returns:

the output size

forwardPropagate

public Matrix forwardPropagate(Matrix x)

Description copied from class: Layer

Forward propagate a batch of input.

Specified by:

forwardPropagate in class Layer

Parameters:

x - input matrix

Returns:

output matrix

getErrors

public Matrix getErrors(Matrix prevErrors)

Description copied from class: Layer

Get backpropagated errors.

Specified by:

getErrors in class Layer

Parameters:

prevErrors - errors from previous layer (layer after output)

Returns:

matrix of errors

getErrorsExpected

public Matrix getErrorsExpected(Matrix y)

Description copied from class: Layer

Get errors from output layer.

Specified by:

getErrorsExpected in class Layer

Parameters:

y - expected outputs

Returns:

matrix of errors

update

public void update(Matrix errors)

Description copied from class: Layer

Update the layer after calculating errors.

Specified by:

update in class Layer

Parameters:

errors - calculated errors

toString

public java.lang.String toString()

Description copied from class: Layer

Get a string representation.

Specified by:

toString in class Layer

Returns:

the string

cloneSettings

public Layer cloneSettings()

Description copied from class: Layer

Clone the settings.

This will return a layer with the same size and activation, but with randomly initialized weights and biases.

Specified by:

cloneSettings in class Layer

Returns:

the clone

cloneLayer

public Layer cloneLayer()

Description copied from class: Layer

Clone the layer, including weights and biases.

Specified by:

cloneLayer in class Layer

Returns:

the clone