wGmdh.jGmdh.oldskul
Class Node

java.lang.Object
  wGmdh.jGmdh.oldskul.Node

All Implemented Interfaces:

java.io.Serializable, java.lang.Comparable<Node>

Direct Known Subclasses:

AttributeNode, Model

public abstract class Node
extends java.lang.Object
implements java.lang.Comparable<Node>, java.io.Serializable

TODO: delegate all data folding and splitting to DatasetHandler ( networkOutput, networkOutputNoFlags)

See Also:

Serialized Form

Nested Class Summary

class

Node.Visited Big cyclic graphs are expected that have significantly less nodes than vertices.

Field Summary

NodeGraphics	graphics
java.util.ArrayList<double[]>	trainSetOutput
java.util.ArrayList<double[]>	validationSetOutput
Node.Visited	visited

Method Summary

int	compareTo(Node n)
static long	getGlobalIdentifier()
long	getIdentifier()
int	getNrFolds()
void	InitializeGraphics(NodeGraphics graphics)
abstract double	networkOutput(double[] inputs, int index) To systematically enable crossvalidation to help decide which structure to go with, data and coefficients learned are organized in ArrayLists - each entry is related to one fold of CV.
abstract double	networkOutputNoFlags(double[] inputs, int index) This one gets invoked by networkOutput.
static void	setGlobalIdentifier(long l)

Methods inherited from class java.lang.Object

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

Field Detail

trainSetOutput

public java.util.ArrayList<double[]> trainSetOutput

validationSetOutput

public java.util.ArrayList<double[]> validationSetOutput

graphics

public transient NodeGraphics graphics

visited

public Node.Visited visited

Method Detail

getNrFolds

public int getNrFolds()
               throws java.lang.Exception

Throws:

java.lang.Exception

setGlobalIdentifier

public static void setGlobalIdentifier(long l)

getGlobalIdentifier

public static long getGlobalIdentifier()

InitializeGraphics

public void InitializeGraphics(NodeGraphics graphics)

getIdentifier

public long getIdentifier()

compareTo

public int compareTo(Node n)

Specified by:

compareTo in interface java.lang.Comparable<Node>

networkOutput

public abstract double networkOutput(double[] inputs,
                                     int index)

To systematically enable crossvalidation to help decide which structure to go with, data and coefficients learned are organized in ArrayLists - each entry is related to one fold of CV. Call this one when you need to calculate entire network

Parameters:

inputs - ...the Model is fed with

index - ... of CV fold

Returns:

output the node gives when given an input

networkOutputNoFlags

public abstract double networkOutputNoFlags(double[] inputs,
                                            int index)

This one gets invoked by networkOutput. Flags are a must, speed issues otherwise.

Parameters:

inputs -

index -

Returns: