Neuron Class Reference

Description

A neuron is a basic Nucleus.

A neuron combines the weighted input from other neurons and applies an activation function to it to compute the output value:

Vector3 combination = NanoBrain::Neuron::Combinator(bias, synapses);
Vector3 output = NanoBrain::Neuron::Activator(combination);

The synapses are connections to other neurons. Each connection has a weight which is used to multiply the output of that other neuron before it is used by the combinator.

Inheritance diagram for Neuron:

Public Types
enum	CombinatorType { Sum , Product }
	The type of combinators. More...
enum	ActivationType {   Linear , Power , Sqrt , Reciprocal ,   Tanh , Binary , Normalized , Custom }
	The type of.
enum	Type { None , Neuron , MemoryCell , Cluster }
	The types of Nucleus.

Public Member Functions
	Neuron (Cluster parent, string name)
	Create a new Neuron in a Cluster instance.
Synapse	AddSynapse (Neuron sendingNucleus, float weight=1)
	Add a new synapse to this nuclues.
Synapse	GetSynapse (Nucleus sender)
	Find a synapse.
void	RemoveSynapse (Nucleus sendingNucleus)
	Remove a synapse from a Nucleus.
virtual void	SetBias (Vector3 inputValue)
	Set the bias, recalculate the output and update all Nuclei receiving from this Nucleus.
void	SleepCheck ()
	Check if the neuron is sleeping.
override Nucleus	ShallowCloneTo (Cluster parent)
	Function to make a partial clone of this nucleus.
override void	UpdateStateIsolated ()
	Update the state without updating other Nuclei.
virtual void	AddReceiver (Nucleus receiverToAdd, float weight=1)
	Add a new receiver to this neuron.
virtual void	RemoveReceiver (Nucleus receiverToRemove)
	Remove a receiver to this neuron.
virtual void	ProcessStimulus (Vector3 inputValue)
	Process an external stimulus.

Static Public Member Functions
static void	Delete (Nucleus nucleus)
	Delete the give neuron.
static float3	CombinatorSum (float3 bias, List< Synapse > synapses)
	Sum the bias and synpase outputs together.
static float3	CombinatorProduct (float3 bias, List< Synapse > synapses)
	Multiply the synapse outputs together.
static Vector3	CombinatorSum (Vector3 bias, List< Synapse > synapses)
	Sum the bias and synpase outputs together.
static Vector3	CombinatorProduct (Vector3 bias, List< Synapse > synapses)
	Multiply the synapse outputs together.

Public Attributes
Vector3	bias = Vector3.zero
	The bias.
CombinatorType	combinator = CombinatorType.Sum
	The type of combinator used for this Neuron.
ActivationType	_activator
	The activation function.
Action	WhenFiring
	An action which is called every time the neuron is updated and is firing.
bool	persistOutput = false
	When true, the value will not be reset after timeToSleep.
float	lastUpdate = 0
	The time at which the last update has been done.
bool	breakOnUpdate = false
	When true, Unity will pause exection when this neuron is updated.
string	name
	The name of the Nucleus.
Cluster	parent
	The cluster instance in which the nucleus is located.

Static Public Attributes
static readonly float	timeToSleep = 0.5f
	Time in seconds after the last update the neuron can go to sleep.

Protected Member Functions
virtual void	CloneFields (Neuron clone)
	Copy relevant fields of this neuron to the given neuron.
float3	Combinator (float3 bias, List< Synapse > synapses)
	The combinator which combines the bias with the values from all synapses.
Vector3	Combinator (Vector3 bias, List< Synapse > synapses)
	The combinator which combines the bias with the values from all synapses.
float3	Activator (float3 inputValue)
	Apply the activation function to the input.
float3	ActivatorLinear (float3 input)
	Linear activation function.
float3	ActivatorSqrt (float3 input)
	Square root activation function.
float3	ActivatorPower (float3 input)
	Power activation function.
float3	ActivatorReciprocal (float3 input)
	Reciprocal activation function.
float3	ActivatorTanh (float3 input)
	Tanh activation function.
float3	ActivatorBinary (float3 input)
	Binary activation function.
float3	ActivatorNormalized (float3 input)
	Normalize activation function.
Vector3	Activator (Vector3 inputValue)
	Apply the activation function to the input.
Vector3	ActivatorLinear (Vector3 input)
	Linear activation function.
Vector3	ActivatorSqrt (Vector3 input)
	Square root activation function.
Vector3	ActivatorPower (Vector3 input)
	Power activation function.
Vector3	ActivatorReciprocal (Vector3 input)
	Reciprocal activation function.

Protected Attributes
float3	_outputValue
	The output value of the neuron.
Vector3	_outputValue
	The output value of the neuron.

Properties
List< Synapse >	synapses [get]
	The synapses of the nucleus.
ActivationType	activator [get, set]
	The activation funtion.
virtual float3	outputValue [get, set]
	The output value of the neuron.
float	outputMagnitude [get]
	The magnitude of the neuron output.
float	outputSqrMagnitude [get]
	The squared magnitude of the neuron output.
virtual Vector3	outputValue [get, set]
	The output value of the neuron.
bool	isFiring [get]
	True if the neuron have a positive value with magnitude > 0.5.
virtual bool	isSleeping [get]
	True when the neuron is not persisting and has not be updated for timeToSleep seconds.
virtual List< Nucleus >	receivers [get, set]
	The nuclei which have a synapse to this neuron.

Member Enumeration Documentation

CombinatorType

enum CombinatorType

The type of combinators.

A combinator combines the weighted values of the synapses to a single value

Enumerator
Sum	Add the weighted values together.
Product	Multiply the weighted values.

Constructor & Destructor Documentation

Neuron()

Neuron	(	Cluster	parent,
		string	name
	)

Create a new Neuron in a Cluster instance.

Parameters

parent	The parent cluster in which the new Neuron should be created
name	The name of the new Neuron

Member Function Documentation

AddSynapse()

Synapse AddSynapse	(	Neuron	sendingNucleus,
		float	weight = 1
	)

Add a new synapse to this nuclues.

Parameters

sendingNucleus	The nucleus from which the signals may originate
weight	The weight applied to the input. Default value = 1

Returns

The created Synapse

This will add a new input to this nucleus with the given weight.

GetSynapse()

Synapse GetSynapse

(

Nucleus

sender

)

Find a synapse.

Parameters

sender

The sender of the input to the Synapse

Returns

The found Synapse or null when the sender has no synapse to this nucleus.

RemoveSynapse()

void RemoveSynapse

(

Nucleus

sendingNucleus

)

Remove a synapse from a Nucleus.

Parameters

sendingNucleus

Remote the synapse connecting to this Nucleus

SetBias()

virtual void SetBias ( Vector3  inputValue )

virtual

Set the bias, recalculate the output and update all Nuclei receiving from this Nucleus.

Parameters

inputValue

SleepCheck()

void SleepCheck

(

)

Check if the neuron is sleeping.

This will reset the output value if it is sleeping

ShallowCloneTo()

override Nucleus ShallowCloneTo

(

Cluster

parent

)

Function to make a partial clone of this nucleus.

Parameters

parent

The cluster in which the cloned nucleus should be placed

Returns

CloneFields()

virtual void CloneFields ( Neuron  clone )

protectedvirtual

Copy relevant fields of this neuron to the given neuron.

Parameters

clone

Delete()

static void Delete ( Nucleus  nucleus )

static

Delete the give neuron.

Parameters

nucleus

The neuron to delete

UpdateStateIsolated()

override void UpdateStateIsolated

(

)

Update the state without updating other Nuclei.

Combinator() [1/2]

float3 Combinator ( float3  bias, List< Synapse >  synapses  )

protected

The combinator which combines the bias with the values from all synapses.

Parameters

bias	The bias of the neuron
synapses	The synapses of the neuron

Returns

CombinatorSum() [1/2]

static float3 CombinatorSum ( float3  bias, List< Synapse >  synapses  )

static

Sum the bias and synpase outputs together.

Parameters

bias	The bias of the neuron
synapses	The synapses of the neuron

Returns

CombinatorProduct() [1/2]

static float3 CombinatorProduct ( float3  bias, List< Synapse >  synapses  )

static

Multiply the synapse outputs together.

Parameters

bias	The bias of the neuron
synapses	The synapses of the neuron

Returns

The result of the multiplication

Combinator() [2/2]

Vector3 Combinator ( Vector3  bias, List< Synapse >  synapses  )

protected

The combinator which combines the bias with the values from all synapses.

Parameters

bias	The bias of the neuron
synapses	The synapses of the neuron

Returns

CombinatorSum() [2/2]

static Vector3 CombinatorSum ( Vector3  bias, List< Synapse >  synapses  )

static

Sum the bias and synpase outputs together.

Parameters

bias	The bias of the neuron
synapses	The synapses of the neuron

Returns

CombinatorProduct() [2/2]

static Vector3 CombinatorProduct ( Vector3  bias, List< Synapse >  synapses  )

static

Multiply the synapse outputs together.

Parameters

bias	The bias of the neuron
synapses	The synapses of the neuron

Returns

The result of the multiplication

Activator() [1/2]

float3 Activator ( float3  inputValue )

protected

Apply the activation function to the input.

Parameters

inputValue

Returns

The result of applying the activation function

ActivatorLinear() [1/2]

float3 ActivatorLinear ( float3  input )

protected

Linear activation function.

Parameters

input

Input value

Returns

The unchanged value

ActivatorSqrt() [1/2]

float3 ActivatorSqrt ( float3  input )

protected

Square root activation function.

Parameters

input

Input value

Returns

The square root of the input

ActivatorPower() [1/2]

float3 ActivatorPower ( float3  input )

protected

Power activation function.

Parameters

input

Input value

Returns

The input to the power of 2

ActivatorReciprocal() [1/2]

float3 ActivatorReciprocal ( float3  input )

protected

Reciprocal activation function.

Parameters

input

Input value

Returns

1/input value

ActivatorTanh()

float3 ActivatorTanh ( float3  input )

protected

Tanh activation function.

Parameters

input

Input value

Returns

Tanh(input value)

ActivatorBinary()

float3 ActivatorBinary ( float3  input )

protected

Binary activation function.

Parameters

input

Input value

Returns

An uniform vector with magnitude between 0 and 1

ActivatorNormalized()

float3 ActivatorNormalized ( float3  input )

protected

Normalize activation function.

Parameters

input

Input value

Returns

The normalized vector

Activator() [2/2]

Vector3 Activator ( Vector3  inputValue )

protected

Apply the activation function to the input.

Parameters

inputValue

Returns

The result of applying the activation function

ActivatorLinear() [2/2]

Vector3 ActivatorLinear ( Vector3  input )

protected

Linear activation function.

Parameters

input

Input value

Returns

The unchanged value

ActivatorSqrt() [2/2]

Vector3 ActivatorSqrt ( Vector3  input )

protected

Square root activation function.

Parameters

input

Input value

Returns

The square root of the input

ActivatorPower() [2/2]

Vector3 ActivatorPower ( Vector3  input )

protected

Power activation function.

Parameters

input

Input value

Returns

The input to the power of 2

ActivatorReciprocal() [2/2]

Vector3 ActivatorReciprocal ( Vector3  input )

protected

Reciprocal activation function.

Parameters

input

Input value

Returns

1/input value

AddReceiver()

virtual void AddReceiver ( Nucleus  receiverToAdd, float  weight = 1  )

virtual

Add a new receiver to this neuron.

Parameters

receiverToAdd	The receiver to add
weight	The weight to use for the synapse to his neuron

RemoveReceiver()

virtual void RemoveReceiver ( Nucleus  receiverToRemove )

virtual

Remove a receiver to this neuron.

Parameters

receiverToRemove

The receiver to remove

ProcessStimulus()

virtual void ProcessStimulus ( Vector3  inputValue )

virtual

Process an external stimulus.

Parameters

inputValue

The value of the stimulus

Member Data Documentation

bias

Vector3 bias = Vector3.zero

The bias.

The bias which a value which is always added to the combined value of the neuron It does not have a synapse and therefore no weight of source nucleus

breakOnUpdate

bool breakOnUpdate = false

When true, Unity will pause exection when this neuron is updated.

Pausing is implemented using Debug.Break()