Layer Class Reference

This is the class for creating layers that are contained inside the Net class. More...

#include <Layer.h>

Public Types
enum	whichGradient { exploding = 0, average = 1, vanishing = 2 }
	Options for what gradient of a chosen error to monitor.

Public Member Functions
	Layer (int _nNeurons, int _nInputs, int _subject, string _trial)
	Constructor for Layer: it initialises the neurons internally. More...
	~Layer ()
	Destructor De-allocated any memory.
void	initLayer (int _layerIndex, Neuron::weightInitMethod _wim, Neuron::biasInitMethod _bim, Neuron::actMethod _am)
	Initialises each layer with specific methods for weight/bias initialisation and activation function of neurons. More...
void	setlearningRate (double _w_learningRate, double _b_learningRate)
	Sets the learning rate. More...
void	setInputs (const double *_inputs, const double scale=1.0, const unsigned int offset=0, const int n=-1)
	Sets the inputs to all neurons in the first hidden layer only. More...
void	propInputs (int _index, double _value)
	Sets the inputs to all neurons in the deeper layers (excluding the first hidden layer) More...
void	calcOutputs ()
	Demands that all neurons in this layer calculate their output.
void	setError (double _error)
	Sets the error to be propagated backward at all neurons in the output layer only. More...
double	getGradient (whichGradient _whichGradient)
	Allows for accessing the error that propagates backward in the network. More...
void	updateWeights ()
	Requests that all neurons perform one iteration of learning.
Neuron *	getNeuron (int _neuronIndex)
	Allows access to a specific neuron. More...
int	getnNeurons ()
	Reports the number of neurons in this layer. More...
double	getOutput (int _neuronIndex)
	Allows for accessing the activation of a specific neuron. More...
double	getSumOutput (int _neuronIndex)
	Allows for accessing the sum output of any specific neuron. More...
double	getWeights (int _neuronIndex, int _weightIndex)
	Allows for accessing any specific weights in the layer. More...
double	getWeightChange ()
	Accesses the total sum of weight changes of all the neurons in this layer. More...
double	getWeightDistance ()
	Performs squared root on the weight change. More...
double	getGlobalError (int _neuronIndex)
	Reports the global error that is assigned to a specific neuron in this layer. More...
double	getInitWeight (int _neuronIndex, int _weightIndex)
	Reports the initial value that was assigned to a specific weight at the initialisatin of the network. More...
void	saveWeights ()
	Saves the temporal weight change of all weights in all neurons into files.
void	snapWeights (string prefix, string _trial, int _subject)
	Snaps the final distribution of weights in a specific layer, this is overwritten every time the function is called.
void	snapWeightsMatrixFormat (string prefix)
void	printLayer ()
	Prints on the console a full tree of this layer with the values of all weights and outputs for all neurons.

Detailed Description

This is the class for creating layers that are contained inside the Net class.

The Layer instances in turn contain neurons.

Constructor & Destructor Documentation

Layer()

Layer::Layer	(	int	_nNeurons,
		int	_nInputs,
		int	_subject,
		string	_trial
	)

Constructor for Layer: it initialises the neurons internally.

Parameters

_nNeurons	Total number of neurons in the layer
_nInputs	Total number of inputs to that layer

Member Function Documentation

getGlobalError()

double Layer::getGlobalError

(

int

_neuronIndex

)

Reports the global error that is assigned to a specific neuron in this layer.

Parameters

_neuronIndex

the neuron index

Returns

the value of the global error

getGradient()

double Layer::getGradient

(

whichGradient

_whichGradient

)

Allows for accessing the error that propagates backward in the network.

Parameters

_neuronIndex

The index from which the error is requested

Returns

Returns the error of the chosen neuron

getInitWeight()

double Layer::getInitWeight	(	int	_neuronIndex,
		int	_weightIndex
	)

Reports the initial value that was assigned to a specific weight at the initialisatin of the network.

Parameters

_neuronIndex	Index of the neuron containing the weight
_weightIndex	Index of the weight

Returns

getNeuron()

Neuron* Layer::getNeuron

(

int

_neuronIndex

)

Allows access to a specific neuron.

Parameters

_neuronIndex

The index of the neuron to access

Returns

A pointer to that neuron

getnNeurons()

int Layer::getnNeurons

(

)

Reports the number of neurons in this layer.

Returns

The total number of neurons in this layer

getOutput()

double Layer::getOutput

(

int

_neuronIndex

)

Allows for accessing the activation of a specific neuron.

Parameters

_neuronIndex

The index of the neuron

Returns

the activation of that neuron

getSumOutput()

double Layer::getSumOutput

(

int

_neuronIndex

)

Allows for accessing the sum output of any specific neuron.

Parameters

_neuronIndex

The index of the neuron to access

Returns

Returns the wighted sum of the inputs to that neuron

getWeightChange()

double Layer::getWeightChange

(

)

Accesses the total sum of weight changes of all the neurons in this layer.

Returns

sum of weight changes all neurons

getWeightDistance()

double Layer::getWeightDistance

(

)

Performs squared root on the weight change.

Returns

The sqr of the weight changes

getWeights()

double Layer::getWeights	(	int	_neuronIndex,
		int	_weightIndex
	)

Allows for accessing any specific weights in the layer.

Parameters

_neuronIndex	The index of the neuron containing that weight
_weightIndex	The index of the input to which that weight is assigned

Returns

Returns the chosen weight

initLayer()

void Layer::initLayer	(	int	_layerIndex,
		Neuron::weightInitMethod	_wim,
		Neuron::biasInitMethod	_bim,
		Neuron::actMethod	_am
	)

Initialises each layer with specific methods for weight/bias initialisation and activation function of neurons.

Parameters

_layerIndex	The index that is assigned to this layer by the Net class
_wim	weights initialisation method, see Neuron::weightInitMethod for different options
_bim	biases initialisation method, see Neuron::biasInitMethod for different options
_am	activation method, see Neuron::actMethod for different options

propInputs()

void Layer::propInputs	(	int	_index,
		double	_value
	)

Sets the inputs to all neurons in the deeper layers (excluding the first hidden layer)

Parameters

_index	The index of the input
_value	The value of the input

setError()

void Layer::setError

(

double

_error

)

Sets the error to be propagated backward at all neurons in the output layer only.

Parameters

_leadError

the error to be propagated backward

setInputs()

void Layer::setInputs	(	const double *	_inputs,
		const double	scale = 1.0,
		const unsigned int	offset = 0,
		const int	n = -1
	)

Sets the inputs to all neurons in the first hidden layer only.

Parameters

_inputs

A pointer to an array of inputs

setlearningRate()

void Layer::setlearningRate	(	double	_w_learningRate,
		double	_b_learningRate
	)

Sets the learning rate.

Parameters

_learningRate

Sets the learning rate for all neurons.

---

The documentation for this class was generated from the following file:

• Layer.h