Package deepnetts.net.layers

Class OutputLayer

java.lang.Object

deepnetts.net.layers.AbstractLayer<TensorBase,TensorBase,Tensor2D>

deepnetts.net.layers.OutputLayer

All Implemented Interfaces:

Layer<TensorBase>, Serializable

Direct Known Subclasses:

SoftmaxOutputLayer

public class OutputLayer
extends AbstractLayer<TensorBase,TensorBase,Tensor2D>

Output layer of a neural network. It is always the last layer in a neural network, and gives the final output of a network.

See Also:

• AbstractLayer
• Layer
• NeuralNetwork
• Serialized Form

Constructor Summary

Constructors

Constructor	Description
OutputLayer(int width)	Creates an instance of output layer with specified width (number of outputs) and sigmoid activation function by default.
OutputLayer(int width, ActivationType actType)	Creates an instance of output layer with specified width (number of outputs) and specified activation function.
OutputLayer(String[] outputLabels, ActivationType actType)	Creates an instance of output layer with specified width (number of outputs) and linear activation function by default.

Method Summary

Modifier and Type	Method	Description
void	applyWeightChanges()	Applies weight changes after one learning iteration or batch
void	backward()	This method implements backward pass for the output layer.
void	forward()	This method implements forward pass for the output layer.
final LossType	getLossType()
final TensorBase	getOutputErrors()
float	getSingleOutInput()
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	setLossType(LossType lossType)
final void	setOutputErrors(TensorBase outputErrors)
String	toString()

Methods inherited from class deepnetts.net.layers.AbstractLayer

getActivation, getActivationType, getBatchSize, getBiases, getDeltaBiases, getDeltas, getDeltaWeights, getDepth, getForwardAccelerator, getGradients, getHeight, getL1Regularization, getL1WeightSum, getL2Regularization, getL2WeightSum, getLearningRate, getMode, getMomentum, getNextLayer, getOptimizer, getOptimizerType, getOutputs, getPrevDeltaBiases, getPrevDeltaWeights, getPrevlayer, getWeights, getWidth, initTransientFields, isBatchMode, isTrainable, setBatchMode, setBatchSize, setBiases, setCudaHandles, setDeltas, setL1Regularization, setL2Regularization, setLearningRate, setMode, setMomentum, setNextlayer, setOptimizerType, setOutputs, setPrevDeltaWeights, setPrevLayer, setThreadPool, setTrainable, setWeights, setWeights

Methods inherited from class java.lang.Object

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

Constructor Details

OutputLayer

public OutputLayer(int width)

Creates an instance of output layer with specified width (number of outputs) and sigmoid activation function by default. Outputs are labeled using generic names "Output1, 2, 3..."

Parameters:

width - layer width which represents number of network outputs

OutputLayer

public OutputLayer(int width,
 ActivationType actType)

Creates an instance of output layer with specified width (number of outputs) and specified activation function. Outputs are labeled using generic names "Output1, 2, 3..."

Parameters:

width - layer width whic represents number of network outputs

actType - activation function

OutputLayer

public OutputLayer(String[] outputLabels,
 ActivationType actType)

Creates an instance of output layer with specified width (number of outputs) and linear activation function by default. Typically linear activation is used for regression tasks, while sigmoid activation is used for binary classification problems.

Parameters:

outputLabels - labels for network's outputs

Method Details

setOutputErrors

public final void setOutputErrors(TensorBase outputErrors)

getOutputErrors

public final TensorBase getOutputErrors()

getLossType

public final LossType getLossType()

setLossType

public void setLossType(LossType lossType)

init

public void init()

Description copied from class: AbstractLayer

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

Specified by:

init in class AbstractLayer<TensorBase,TensorBase,Tensor2D>

forward

public void forward()

This method implements forward pass for the output layer. Calculates weighted input and layer outputs using sigmoid function.

Specified by:

forward in interface Layer<TensorBase>

Specified by:

forward in class AbstractLayer<TensorBase,TensorBase,Tensor2D>

backward

public void backward()

This method implements backward pass for the output layer. http://peterroelants.github.io/posts/neural_network_implementation_intermezzo01/ http://neuralnetworksanddeeplearning.com/chap3.html#introducing_the_cross-entropy_cost_function http://neuralnetworksanddeeplearning.com/chap3.html

Specified by:

backward in interface Layer<TensorBase>

Specified by:

backward in class AbstractLayer<TensorBase,TensorBase,Tensor2D>

applyWeightChanges

public void applyWeightChanges()

Applies weight changes after one learning iteration or batch

Specified by:

applyWeightChanges in class AbstractLayer<TensorBase,TensorBase,Tensor2D>

toString

public String toString()

Overrides:

toString in class Object

getSingleOutInput

public float getSingleOutInput()