This is the class for creating neurons inside the Layer class. More...

#include <Neuron.h>

Public Types
enum	biasInitMethod { B_NONE = 0, B_RANDOM = 1 }
	Options for method of initialising biases 0 for initialising all weights to zero 1 for initialising all weights to one 2 for initialising all weights to a random value between 0 and 1.

enum	weightInitMethod { W_ZEROS = 0, W_ONES = 1, W_RANDOM = 2, W_ONES_NORM = 3, W_RANDOM_NORM = 5 }
	Options for method of initialising weights 0 for initialising all weights to zero 1 for initialising all weights to one 2 for initialising all weights to a random value between 0 and 1.

enum	actMethod { Act_Sigmoid = 1, Act_Tanh = 2, Act_ReLU = 3, Act_NONE = 0 }
	Options for activation functions of the neuron 0 for using the logistic function 1 for using the hyperbolic tan function 2 for unity function (no activation)

enum	whichError { onBackwardError = 0, onMidError = 1, onForwardError = 2 }
	Options for choosing an error to monitor the gradient of 0 for monitoring the error that propagates backward 1 for monitoring the error that propagates from the middle and bilaterally 2 for monitoring the error that propagates forward.

Public Member Functions
	Neuron (int _nInputs)
	Constructor for the Neuron class: it initialises a neuron with specific number fo inputs to that neuron. More...

	~Neuron ()
	Destructor De-allocated any memory.

void	initNeuron (int _neuronIndex, int _layerIndex, weightInitMethod _wim, biasInitMethod _bim, actMethod _am)
	Initialises the neuron with the given methods for weight/bias initialisation and for activation function. More...

void	setLearningRate (double _learningRate, double _b_learningRate)
	Sets the learning rate. More...

void	setInput (int _index, double _value)
	Sets the inputs to this neuron that is located in the first hidden layer. More...

void	propInputs (int _index, double _value)
	Sets the inputs to this neuron that can be located in any layer other than the first hidden layer. More...

int	calcOutput (int _layerHasReported)
	Calculates the output of the neuron by performing a weighed sum of all inputs to this neuron and activating the sum. More...

void	setError (double _value)
	Sets the error of the neuron in the first hidden layer that is to be propagated forward. More...

double	getError ()
	Allows accessing the error of this neuron. More...

void	updateWeights ()
	Performs one iteration of learning, that is: it updates all the weights assigned to each input to this neuron.

double	doActivation (const double sum) const
	Performs the activation of the sum output of the neuron. More...

double	doActivationPrime (const double input) const
	Performs inverse activation on any input that is passed to this function. More...

void	setBackpropError (const double upstreamDeltaErrorSum)
	Sets the internal backprop error. More...

double	getOutput ()
	Requests the output of this neuron. More...

double	getSumOutput ()
	Requests the sum output of the neuron. More...

double	getWeights (int _inputIndex)
	Requests a specific weight. More...

double	getInitWeights (int _inputIndex)
	Requests a inital value of a specific weight. More...

double	getWeightChange ()
	Requests for overall change of all weights contained in this neuron. More...

double	getMaxWeight ()
	Requests for the maximum weights located in this neuron. More...

double	getMinWeight ()
	Requests for the minimum weights located in this neuron. More...

double	getSumWeight ()
	Requests for the total sum of weights located in this neuron. More...

double	getWeightDistance ()
	Requests the weight distance of all weighs in this neuron. More...

int	getnInputs ()
	Requests the total number of inputs to this neuron. More...

void	saveWeights ()
	Saves the temporal weight change of all weights in this neuron into a file.

void	printNeuron ()
	Prints on the console a full description of all weights, inputs and outputs for this neuron.

void	setWeight (int _index, double _weight)
	Sets the weights of the neuron. More...

Detailed Description

This is the class for creating neurons inside the Layer class.

This is the building block class of the network.

Constructor & Destructor Documentation

◆ Neuron()

Neuron::Neuron ( int _nInputs )

Constructor for the Neuron class: it initialises a neuron with specific number fo inputs to that neuron.

Parameters

_nInputs

Member Function Documentation

◆ calcOutput()

int Neuron::calcOutput ( int _layerHasReported )

Calculates the output of the neuron by performing a weighed sum of all inputs to this neuron and activating the sum.

Parameters

_layerHasReported boolean variable to indicate whether or not any neuron in this layer has reported exploding output

Returns: Returns a boolean to report whether or not this neuron has exploding output

◆ doActivation()

double Neuron::doActivation ( const double sum ) const

inline

Performs the activation of the sum output of the neuron.

Parameters

_sum	the weighted sum of all inputs

Returns: activation of the sum

◆ doActivationPrime()

double Neuron::doActivationPrime ( const double input ) const

inline

Performs inverse activation on any input that is passed to this function.

Parameters

_input the input value

Returns: the inverse activation of the input

◆ getError()

double Neuron::getError ( )

Allows accessing the error of this neuron.

Returns: the value of the error

◆ getInitWeights()

double Neuron::getInitWeights ( int _inputIndex )

Requests a inital value of a specific weight.

Parameters

_inputIndex index of the input to which the weight is assigned

Returns: teh inital value of the weight

◆ getMaxWeight()

double Neuron::getMaxWeight ( )

inline

Requests for the maximum weights located in this neuron.

Returns: Returns the max weight

◆ getMinWeight()

double Neuron::getMinWeight ( )

inline

Requests for the minimum weights located in this neuron.

Returns: Returns the min weight

◆ getnInputs()

int Neuron::getnInputs ( )

Requests the total number of inputs to this neuron.

Returns: total number of inputs

◆ getOutput()

double Neuron::getOutput ( )

inline

Requests the output of this neuron.

Returns: the output of the neuron after activation

◆ getSumOutput()

double Neuron::getSumOutput ( )

inline

Requests the sum output of the neuron.

Returns: returns the sum output of the neuron before activaiton

◆ getSumWeight()

double Neuron::getSumWeight ( )

inline

Requests for the total sum of weights located in this neuron.

Returns: Returns the sum of weights

◆ getWeightChange()

double Neuron::getWeightChange ( )

Requests for overall change of all weights contained in this neuron.

Returns: the overal weight change

◆ getWeightDistance()

double Neuron::getWeightDistance ( )

Requests the weight distance of all weighs in this neuron.

Returns: returns the sqr of the total weight change in this neuron

◆ getWeights()

double Neuron::getWeights ( int _inputIndex )

Requests a specific weight.

Parameters

_inputIndex index of the input to which the chosen weight is assigned

Returns: Returns the chosen weight

◆ initNeuron()

void Neuron::initNeuron	(	int	_neuronIndex,
		int	_layerIndex,
		weightInitMethod	_wim,
		biasInitMethod	_bim,
		actMethod	_am
	)

Initialises the neuron with the given methods for weight/bias initialisation and for activation function.

It also specifies the index of the neuron and the index of the layer that contains this neuron.

Parameters

_neuronIndex	The index of this neuron
_layerIndex	The index of the layer that contains this neuron
_wim	The method of initialising the weights, refer to weightInitMethod for more information
_bim	The method of initialising the biases, refer to biasInitMethod for more information
_am	The function used for activation of neurons, refer to actMethod for more information

◆ propInputs()

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

Sets the inputs to this neuron that can be located in any layer other than the first hidden layer.

Parameters

_index	index of the input
_value	value of the input

◆ setBackpropError()

void Neuron::setBackpropError ( const double upstreamDeltaErrorSum )

inline

Sets the internal backprop error.

Parameters

_input the input value

◆ setError()

void Neuron::setError ( double _value )

Sets the error of the neuron in the first hidden layer that is to be propagated forward.

Parameters

_value value of the error

◆ setInput()

void Neuron::setInput	(	int	_index,
		double	_value
	)

Sets the inputs to this neuron that is located in the first hidden layer.

Parameters

_index	Index of the input
_value	Value of the input

◆ setLearningRate()

void Neuron::setLearningRate	(	double	_learningRate,
		double	_b_learningRate
	)

Sets the learning rate.

Parameters

_learningRate Sets the learning rate for this neuron.

◆ setWeight()

void Neuron::setWeight	(	int	_index,
		double	_weight
	)

inline

Sets the weights of the neuron.

Parameters

_index	index of the weight
_weight	value of the weight

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

Neuron.h

Public Types

Public Member Functions

Detailed Description

Constructor & Destructor Documentation

◆ Neuron()

Member Function Documentation

◆ calcOutput()

◆ doActivation()

◆ doActivationPrime()

◆ getError()

◆ getInitWeights()

◆ getMaxWeight()

◆ getMinWeight()

◆ getnInputs()

◆ getOutput()

◆ getSumOutput()

◆ getSumWeight()

◆ getWeightChange()

◆ getWeightDistance()

◆ getWeights()

◆ initNeuron()

◆ propInputs()

◆ setBackpropError()

◆ setError()

◆ setInput()

◆ setLearningRate()

◆ setWeight()