org.aika.lattice

Class Node<T extends Node,A extends NodeActivation<T>>

java.lang.Object

org.aika.AbstractNode<Provider<T>>

org.aika.lattice.Node<T,A>

All Implemented Interfaces:

Comparable<Node>, Writable

Direct Known Subclasses:

AndNode, InputNode, OrNode

public abstract class Node<T extends Node,A extends NodeActivation<T>>
extends AbstractNode<Provider<T>>
implements Comparable<Node>

The Node class is the abstract class for all the boolean logic nodes underneath the neural network layer. These nodes form a boolean representation for all the neurons of the neural network. Whenever changes occur to the synapse weights in the neural layer, then the structure of the boolean representation needs to be adjusted. Several neurons, however, might share common substructures in this boolean representation. The InputNode and the AndNode classes together form a pattern lattice, containing all possible substructures of any given conjunction. For example if we have the conjunction ABCD where A, B, C, D are the inputs then the pattern lattice will contain the nodes ABCD, ABC, ABD, ACD, BCD, AB, AC, AD, BC, BD, CD, A, B, C, D. The class OrNode is a disjunction of either input-nodes or and-nodes. The or-node is connected with one of the neurons.

Each logic node has a set of activations. The activations are stored in the thread local data structure ThreadState.

Author:

Lukas Molzberger

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	Node.RidVisited Aika extensively uses graph coloring techniques.
static class	Node.ThreadState<T extends Node,A extends NodeActivation<T>> The ThreadState is a thread local data structure containing the activations of a single document for a specific logic node.

Field Summary

Fields

Modifier and Type	Field and Description
TreeSet<org.aika.lattice.OrNode.OrEntry>	allOrChildren
TreeMap<AndNode.Refinement,Provider<AndNode>>	andChildren
static Comparator<NodeActivation.Key>	BEGIN_COMP
static Comparator<NodeActivation.Key>	END_COMP
boolean	isDiscovered
int	level
ReadWriteLock	lock
static Node	MAX_NODE
static int	MAX_RELATIVE_RID
static Node	MIN_NODE
AtomicInteger	numberOfNeuronRefs
TreeSet<org.aika.lattice.OrNode.OrEntry>	orChildren
TreeMap<org.aika.lattice.Node.ReverseAndRefinement,AndNode.Refinement>	reverseAndChildren
static Comparator<NodeActivation.Key>	RID_COMP
Writable	statistic
Node.ThreadState<T,A>[]	threads

Fields inherited from class org.aika.AbstractNode

lastUsedDocumentId, provider

Constructor Summary

Constructors

Modifier	Constructor and Description
protected	Node()
	Node(Model m, int level)

Method Summary

Modifier and Type	Method and Description
static <T extends Node,A extends NodeActivation<T>> void	addActivationAndPropagate(Document doc, NodeActivation.Key<T> ak, Collection<NodeActivation<?>> inputActs) Add a new activation to this logic node and further propagate this activation through the network.
void	changeNumberOfNeuronRefs(int threadId, long v, int d)
abstract void	cleanup()
int	compareTo(Node n)
abstract double	computeSynapseWeightSum(Integer offset, INeuron n)
abstract void	discover(Document doc, NodeActivation<T> act, PatternDiscovery.Config config)
String	getNeuronLabel()
Node.ThreadState<T,A>	getThreadState(int threadId, boolean create)
abstract boolean	isAllowedOption(int threadId, InterpretationNode n, NodeActivation<?> act, long v)
boolean	isRequired()
abstract String	logicToString()
void	processChanges(Document doc) Process all added or removed activation for this logic node.
abstract void	propagateAddedActivation(Document doc, A act) Propagate an activation to the next node or the next neuron that is depending on the current node.
void	readFields(DataInput in, Model m) Deserialize the fields of this object from in.
static Node	readNode(DataInput in, Provider p)
void	register(A act, Document doc)
void	remove()
String	toString()
String	weightsToString()
void	write(DataOutput out) Serialize the fields of this object to out.

Methods inherited from class org.aika.AbstractNode

reactivate, read, setModified, suspend

Methods inherited from class java.lang.Object

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

Field Detail

MAX_RELATIVE_RID

public static int MAX_RELATIVE_RID

MIN_NODE

public static final Node MIN_NODE

MAX_NODE

public static final Node MAX_NODE

reverseAndChildren

public TreeMap<org.aika.lattice.Node.ReverseAndRefinement,AndNode.Refinement> reverseAndChildren

andChildren

public TreeMap<AndNode.Refinement,Provider<AndNode>> andChildren

orChildren

public TreeSet<org.aika.lattice.OrNode.OrEntry> orChildren

allOrChildren

public TreeSet<org.aika.lattice.OrNode.OrEntry> allOrChildren

level

public int level

statistic

public Writable statistic

isDiscovered

public boolean isDiscovered

numberOfNeuronRefs

public AtomicInteger numberOfNeuronRefs

lock

public ReadWriteLock lock

threads

public Node.ThreadState<T extends Node,A extends NodeActivation<T>>[] threads

BEGIN_COMP

public static final Comparator<NodeActivation.Key> BEGIN_COMP

END_COMP

public static final Comparator<NodeActivation.Key> END_COMP

RID_COMP

public static final Comparator<NodeActivation.Key> RID_COMP

Constructor Detail

Node

protected Node()

Node

public Node(Model m,
            int level)

Method Detail

getThreadState

public Node.ThreadState<T,A> getThreadState(int threadId,
                                            boolean create)

propagateAddedActivation

public abstract void propagateAddedActivation(Document doc,
                                              A act)

Propagate an activation to the next node or the next neuron that is depending on the current node.

Parameters:

doc -

act -

isAllowedOption

public abstract boolean isAllowedOption(int threadId,
                                        InterpretationNode n,
                                        NodeActivation<?> act,
                                        long v)

computeSynapseWeightSum

public abstract double computeSynapseWeightSum(Integer offset,
                                               INeuron n)

discover

public abstract void discover(Document doc,
                              NodeActivation<T> act,
                              PatternDiscovery.Config config)

cleanup

public abstract void cleanup()

logicToString

public abstract String logicToString()

register

public void register(A act,
                     Document doc)

processChanges

public void processChanges(Document doc)

Process all added or removed activation for this logic node.

Parameters:

doc -

addActivationAndPropagate

public static <T extends Node,A extends NodeActivation<T>> void addActivationAndPropagate(Document doc,
                                                                                          NodeActivation.Key<T> ak,
                                                                                          Collection<NodeActivation<?>> inputActs)

Add a new activation to this logic node and further propagate this activation through the network. This activation, however, will not be added immediately. This method only adds a request to the activations queue in the document. The activation will be added when the method Node.processChanges(Document doc) is called.

Parameters:

doc -

ak -

inputActs -

remove

public void remove()

isRequired

public boolean isRequired()

changeNumberOfNeuronRefs

public void changeNumberOfNeuronRefs(int threadId,
                                     long v,
                                     int d)

getNeuronLabel

public String getNeuronLabel()

toString

public String toString()

Overrides:

toString in class Object

weightsToString

public String weightsToString()

compareTo

public int compareTo(Node n)

Specified by:

compareTo in interface Comparable<Node>

write

public void write(DataOutput out)
           throws IOException

Description copied from interface: Writable

Serialize the fields of this object to out.

Specified by:

write in interface Writable

Parameters:

out - DataOuput to serialize this object into.

Throws:

IOException

readFields

public void readFields(DataInput in,
                       Model m)
                throws IOException

Description copied from interface: Writable

Deserialize the fields of this object from in.

For efficiency, implementations should attempt to re-use storage in the existing object where possible.

Specified by:

readFields in interface Writable

Parameters:

in - DataInput to deseriablize this object from.

Throws:

IOException

readNode

public static Node readNode(DataInput in,
                            Provider p)
                     throws IOException

Throws:

IOException