Neuron Class Reference

#include <Neuron.h>

List of all members.

Public Member Functions
	Neuron (int numInputs) throw (std::bad_alloc)
	Neuron (const Neuron &neuron) throw (std::bad_alloc)
Neuron &	operator= (const Neuron &neuron) throw (std::bad_alloc)
void	SetNumInputs (int num)
int	GetNumInputs () const
void	SetActivationFunc (ACTIVATION_FUNC activ_func)
void	SetDerivActivationFunc (DERIV_ACTIVATION_FUNC deriv)
void	SetWeight (int indx, double weight)
void	SetWeights (const std::vector< double > &weights)
double	GetWeight (int indx) const
void	GetWeights (std::vector< double > &weights) const
double *	GetWeights ()
double	GetOutput (const std::vector< double > &inputs, const int *indices)
double	GetOutput (const std::vector< double > &inputs, const int *indices, double &net_input)
double	GetOutput (const double *inputs, const int *indices)
double	GetOutput (const double *inputs, const int *indices, double &net_input)
void	SetIndex (int i)
int	GetIndex () const
void	Clear ()
Public Attributes
ACTIVATION_FUNC	pActivationFunc
DERIV_ACTIVATION_FUNC	pDerivActivationFunc

---

Detailed Description

Neuron class This class describes the representation of neuron: the basic building block of neural networks. It consists of a set of inputs (which can be hooked to other neurons), and corresponding weights for the inputs. The bias should be included in the weights, and the corresponding input set to -1. The neuron computes its output and returns that from GetOutput. Each call to GetOutput will recompute the output. So if the inputs to neuron have not changed since the last call, do not make repeated calls to GetOutput for efficiency. Instead, store it somewhere and reuse. For instance, NeuralNetwork class stores the output from neurons because it needs to use this output of input neurons to copute activations and outputs of hidden neurons.

---

Constructor & Destructor Documentation

Neuron::Neuron (  int  numInputs  )  throw (std::bad_alloc)

Constructs the neuron with specified number of input connections. numInputs: Number of inputs to the neuron

---

Member Function Documentation

void Neuron::Clear (   )

Deallocates space for weights of the neuron and clears up member variables so that neuron can be redefined using SetNumInputs and SetWeights

double* Neuron::GetWeights (   )  [inline]

the function below should not be const because it is used by calibrators to update neuron weights

void Neuron::SetNumInputs (  int  num  )

Set number of inputs to neuron

void Neuron::SetWeight (  int  indx, double  weight )

Gets the weight for specified input connection. Does not validate bounds. indx >= 0 and indx < iNumFrom

---

Member Data Documentation

ACTIVATION_FUNC Neuron::pActivationFunc

Activation function pointer If all neurons in the network use same activation function, compile this class with SAME_ACTIVATION flag defined. This will save space (will not allocate different pointers to funtions for different neurons.

DERIV_ACTIVATION_FUNC Neuron::pDerivActivationFunc

Derivative of activation function

---

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

• Neuron.h
• Neuron.cpp

---