deepnetts.net.layers

Class AbstractLayer

java.lang.Object

deepnetts.net.layers.AbstractLayer

All Implemented Interfaces:

Layer, Serializable

Direct Known Subclasses:

ConvolutionalLayer, FlattenLayer, FullyConnectedLayer, InputLayer, MaxPoolingLayer, OutputLayer

public abstract class AbstractLayer
extends Object
implements Layer, Serializable

Base class for different types of layers. Provides common functionality for all type of layers: layer dimensions, inputs, outputs, connection to previous and/or next layer, activation function and abstract methods for initialization, forward and backward pass.

Author:

Zoran Sevarac

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
AbstractLayer(ActivationType activationType)

Method Summary

Modifier and Type	Method and Description
abstract void	applyWeightChanges() Applies weight changes to current weights Must be diferent for convolutional does nothing for MaxPooling Same for FullyConnected and OutputLayer
abstract void	backward() This method should implement backward pass in subclasses
abstract void	forward() This method should implement forward pass in subclasses
ActivationFunction	getActivation()
ActivationType	getActivationType()
float	getBatchSize()
float[]	getBiases()
float[]	getDeltaBiases()
Tensor	getDeltas() Returns layer deltas/errors (as a tensor).
Tensor	getDeltaWeights()
int	getDepth()
Tensor	getGradients()
int	getHeight()
float	getL1Regularization()
float	getL1WeightSum()
float	getL2Regularization()
float	getL2WeightSum()
float	getLearningRate()
float	getMomentum()
AbstractLayer	getNextLayer()
Optimizer	getOptimizer()
OptimizerType	getOptimizerType()
Tensor	getOutputs() Returns layer outputs (as a tensor).
float[]	getPrevDeltaBiases()
Tensor	getPrevDeltaWeights()
AbstractLayer	getPrevlayer()
Tensor	getWeights()
int	getWidth()
abstract void	init() This method should implement layer initialization in subclasses, when a layer is added to the network (create weights, outputs, deltas, randomization etc.).
void	initTransientFields()
boolean	isBatchMode()
boolean	isTrainable()
void	setBatchMode(boolean batchMode)
void	setBatchSize(int batchSize)
void	setBiases(float[] biases)
void	setDeltas(Tensor deltas)
void	setL1Regularization(float regL1)
void	setL2Regularization(float regL2)
void	setLearningRate(float learningRate)
void	setMomentum(float momentum)
void	setNextlayer(AbstractLayer nextlayer)
void	setOptimizerType(OptimizerType optType)
void	setOutputs(Tensor outputs)
void	setPrevDeltaWeights(Tensor prevDeltaWeights)
void	setPrevLayer(AbstractLayer prevLayer)
void	setTrainable(boolean trainable) Set trainable to false to freeze learned weights.
void	setWeights(String weightStr)
void	setWeights(Tensor weights)

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

AbstractLayer

public AbstractLayer(ActivationType activationType)

Method Detail

init

public abstract void init()

This method should implement layer initialization in subclasses, when a layer is added to the network (create weights, outputs, deltas, randomization etc.).

forward

public abstract void forward()

This method should implement forward pass in subclasses

Specified by:

forward in interface Layer

backward

public abstract void backward()

This method should implement backward pass in subclasses

Specified by:

backward in interface Layer

applyWeightChanges

public abstract void applyWeightChanges()

Applies weight changes to current weights Must be diferent for convolutional does nothing for MaxPooling Same for FullyConnected and OutputLayer

getWidth

public int getWidth()

getHeight

public int getHeight()

getDepth

public int getDepth()

getPrevlayer

public AbstractLayer getPrevlayer()

setPrevLayer

public void setPrevLayer(AbstractLayer prevLayer)

setNextlayer

public void setNextlayer(AbstractLayer nextlayer)

getNextLayer

public AbstractLayer getNextLayer()

getWeights

public Tensor getWeights()

getBiases

public float[] getBiases()

setBiases

public void setBiases(float[] biases)

getOutputs

public final Tensor getOutputs()

Description copied from interface: Layer

Returns layer outputs (as a tensor).

Specified by:

getOutputs in interface Layer

Returns:

layer output as a tensor

getDeltas

public final Tensor getDeltas()

Description copied from interface: Layer

Returns layer deltas/errors (as a tensor). Deltas are accumulated errors propagated from the next layer.

Specified by:

getDeltas in interface Layer

Returns:

layer deltas tensor

getGradients

public final Tensor getGradients()

getDeltaWeights

public Tensor getDeltaWeights()

getPrevDeltaWeights

public Tensor getPrevDeltaWeights()

setPrevDeltaWeights

public void setPrevDeltaWeights(Tensor prevDeltaWeights)

getPrevDeltaBiases

public float[] getPrevDeltaBiases()

getDeltaBiases

public float[] getDeltaBiases()

setOutputs

public final void setOutputs(Tensor outputs)

setWeights

public void setWeights(Tensor weights)

setWeights

public void setWeights(String weightStr)

setDeltas

public final void setDeltas(Tensor deltas)

getActivation

public ActivationFunction getActivation()

getOptimizer

public Optimizer getOptimizer()

getLearningRate

public float getLearningRate()

setLearningRate

public void setLearningRate(float learningRate)

isBatchMode

public boolean isBatchMode()

setBatchMode

public void setBatchMode(boolean batchMode)

getBatchSize

public float getBatchSize()

setBatchSize

public void setBatchSize(int batchSize)

setMomentum

public void setMomentum(float momentum)

getMomentum

public float getMomentum()

getOptimizerType

public OptimizerType getOptimizerType()

setOptimizerType

public void setOptimizerType(OptimizerType optType)

getActivationType

public ActivationType getActivationType()

getL1WeightSum

public float getL1WeightSum()

getL2WeightSum

public float getL2WeightSum()

getL2Regularization

public float getL2Regularization()

setL2Regularization

public void setL2Regularization(float regL2)

getL1Regularization

public float getL1Regularization()

setL1Regularization

public void setL1Regularization(float regL1)

isTrainable

public boolean isTrainable()

setTrainable

public void setTrainable(boolean trainable)

Set trainable to false to freeze learned weights.

Parameters:

trainable -

initTransientFields

public void initTransientFields()