network.aika.neuron.activation

Class Activation

java.lang.Object

network.aika.lattice.NodeActivation<OrNode>

network.aika.lattice.OrNode.OrActivation

network.aika.neuron.activation.Activation

All Implemented Interfaces:

Comparable<NodeActivation<OrNode>>

public final class Activation
extends OrNode.OrActivation

The Activation class is the most central class in Aika. On the one hand it stores the activation value for a given neuron in the State substructure. On the other hand it specifies where this activation is located within the document and to which interpretation it belongs. The Activation.Key therefore consists of the logic node to which this activation belongs. If this logic node is an or-node, then this activation automatically also belongs to the neuron as well. Furthermore, the key contains the char range within the document and the relational id (rid). The relational id might be used to store the word pos for instance. Lastly, the key contain the interpretation node of this activation, specifying to which interpretation this activation belongs.

The activations are linked to each other on two levels. The fields inputs and outputs contain the activation links within the logic layer. The fields neuronInputs and neuronOutputs contain the links on the neural layer.

Author:

Lukas Molzberger

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	Activation.Builder
static class	Activation.Link The SynapseActivation mirror the synapse link in the network of activations.
static class	Activation.Mode
static class	Activation.Rounds Since Aika is a recurrent neural network, it is necessary to compute several rounds of activation values.
static class	Activation.State A State object contains the activation value of an activation object that belongs to a neuron.
class	Activation.StateChange The StateChange class is used to store the state change of an activation that occurs in each node of the binary search tree.

Field Summary

Fields

Modifier and Type	Field and Description
static Comparator<Activation>	ACTIVATION_ID_COMP
Activation.State	avgState
Candidate	candidate
Activation.StateChange	currentStateChange
long	currentStateV
SearchNode.Decision	decision
double	errorSignal
SearchNode.Decision	finalDecision
Activation.Rounds	finalRounds
SearchNode.Decision	inputDecision
Double	inputValue
boolean	isQueued
double	lowerBound
long	markedAncestor
long	markedDirty
long	markedHasCandidate
long	markedPredecessor
static Activation	MAX_ACTIVATION
static int	MAX_PREDECESSOR_DEPTH
static int	MAX_SELF_REFERENCING_DEPTH
static Activation	MIN_ACTIVATION
TreeMap<Activation.Link,Activation.Link>	neuronInputs
TreeMap<Activation.Link,Activation.Link>	neuronOutputs
long	queueId
Range	range
Activation.Rounds	rounds
Map<Integer,Activation.State>	searchStates
TreeSet<Activation.Link>	selectedNeuronInputs
Integer	sequence
Double	targetValue
boolean	ubQueued
double	upperBound

Fields inherited from class network.aika.lattice.OrNode.OrActivation

inputs

Fields inherited from class network.aika.lattice.NodeActivation

doc, id, node, outputsToAndNode, outputsToOrNode, outputToInputNode, repropagateV, visited

Constructor Summary

Constructors

Constructor and Description
Activation(int id, Document doc, OrNode n)
Activation(int id, Document doc, Range r, OrNode n)

Method Summary

Modifier and Type	Method and Description
void	addSynapseActivation(Linker.Direction dir, Activation.Link sa)
void	adjustSelectedNeuronInputs(SearchNode.Decision d)
boolean	checkSelfReferencing(boolean onlySelected, int depth, long v)
void	computeBounds()
Activation.State	computeValueAndWeight(int round)
Collection<Activation>	getConflicts()
List<Activation.Link>	getFinalInputActivationLinks()
List<Activation.Link>	getFinalOutputActivationLinks()
Activation.State	getFinalState()
INeuron	getINeuron()
String	getLabel()
Neuron	getNeuron()
Integer	getSequence()
Activation	getTarget()
String	getText()
boolean	isFinalActivation()
String	linksToString()
void	markDirty(long v)
void	markPredecessor(long v, int depth)
double	process(SearchNode sn, int round, long v)
void	processBounds()
void	saveNewState()
void	saveOldState(Map<Activation,Activation.StateChange> changes, long v)
void	setDecision(SearchNode.Decision newDecision, long v)
void	setTargetValue(Double targetValue)
String	toString()
String	toString(boolean finalOnly, boolean withTextSnippet, boolean withLogic)

Methods inherited from class network.aika.lattice.OrNode.OrActivation

getInputActivation, link

Methods inherited from class network.aika.lattice.NodeActivation

compareTo

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Field Detail

ACTIVATION_ID_COMP

public static final Comparator<Activation> ACTIVATION_ID_COMP

MAX_SELF_REFERENCING_DEPTH

public static int MAX_SELF_REFERENCING_DEPTH

MAX_PREDECESSOR_DEPTH

public static int MAX_PREDECESSOR_DEPTH

MIN_ACTIVATION

public static Activation MIN_ACTIVATION

MAX_ACTIVATION

public static Activation MAX_ACTIVATION

range

public Range range

selectedNeuronInputs

public TreeSet<Activation.Link> selectedNeuronInputs

neuronInputs

public TreeMap<Activation.Link,Activation.Link> neuronInputs

neuronOutputs

public TreeMap<Activation.Link,Activation.Link> neuronOutputs

sequence

public Integer sequence

upperBound

public double upperBound

lowerBound

public double lowerBound

avgState

public Activation.State avgState

searchStates

public Map<Integer,Activation.State> searchStates

rounds

public Activation.Rounds rounds

finalRounds

public Activation.Rounds finalRounds

ubQueued

public boolean ubQueued

isQueued

public boolean isQueued

queueId

public long queueId

markedHasCandidate

public long markedHasCandidate

currentStateV

public long currentStateV

currentStateChange

public Activation.StateChange currentStateChange

markedDirty

public long markedDirty

markedPredecessor

public long markedPredecessor

errorSignal

public double errorSignal

targetValue

public Double targetValue

inputValue

public Double inputValue

inputDecision

public SearchNode.Decision inputDecision

decision

public SearchNode.Decision decision

finalDecision

public SearchNode.Decision finalDecision

candidate

public Candidate candidate

markedAncestor

public long markedAncestor

Constructor Detail

Activation

public Activation(int id,
                  Document doc,
                  OrNode n)

Activation

public Activation(int id,
                  Document doc,
                  Range r,
                  OrNode n)

Method Detail

setTargetValue

public void setTargetValue(Double targetValue)

getLabel

public String getLabel()

getText

public String getText()

getINeuron

public INeuron getINeuron()

getNeuron

public Neuron getNeuron()

addSynapseActivation

public void addSynapseActivation(Linker.Direction dir,
                                 Activation.Link sa)

process

public double process(SearchNode sn,
                      int round,
                      long v)

computeValueAndWeight

public Activation.State computeValueAndWeight(int round)

processBounds

public void processBounds()

computeBounds

public void computeBounds()

getFinalInputActivationLinks

public List<Activation.Link> getFinalInputActivationLinks()

getFinalOutputActivationLinks

public List<Activation.Link> getFinalOutputActivationLinks()

getConflicts

public Collection<Activation> getConflicts()

adjustSelectedNeuronInputs

public void adjustSelectedNeuronInputs(SearchNode.Decision d)

getTarget

public Activation getTarget()

checkSelfReferencing

public boolean checkSelfReferencing(boolean onlySelected,
                                    int depth,
                                    long v)

setDecision

public void setDecision(SearchNode.Decision newDecision,
                        long v)

isFinalActivation

public boolean isFinalActivation()

getFinalState

public Activation.State getFinalState()

getSequence

public Integer getSequence()

markDirty

public void markDirty(long v)

markPredecessor

public void markPredecessor(long v,
                            int depth)

toString

public String toString()

Overrides:

toString in class Object

toString

public String toString(boolean finalOnly,
                       boolean withTextSnippet,
                       boolean withLogic)

linksToString

public String linksToString()

saveOldState

public void saveOldState(Map<Activation,Activation.StateChange> changes,
                         long v)

saveNewState

public void saveNewState()