neuroidnet.phasesegregator
Class StatisticalInputSequence

java.lang.Object
  |
  +--neuroidnet.ntr.Peripheral
        |
        +--neuroidnet.phasesegregator.Peripheral
              |
              +--neuroidnet.phasesegregator.StatisticalInputSequence

All Implemented Interfaces:

Expressive, Serializable

public class StatisticalInputSequence

extends Peripheral

Creates a random input sequence out of possible input patterns (objects). Designed to be instantiated separately from the Network (not in build() anymore).

Also evaluates the network after training to check for the quality of correct and spurious concepts memorized.

TODO: get only one percept from I1 and lower M1's threshold to allow recruitment.

Created: Fri Oct 18 14:12:22 2002

Modified: $Date: 2003/03/17 20:41:56 $

Version:

$Revision: 1.4 $ for this file.

Author:

Cengiz Gunay

See Also:

Serialized Form

Nested Class Summary

(package private) class

StatisticalInputSequence.SimObject

Field Summary

(package private) int	numberOfItemsPerArea
(package private) int	numberOfMedialAreas
(package private) int	numberOfNeuroids
(package private) int	numberOfObjects Can be manipulated by beanshell scripts under scripts/.
(package private) int	replication
(package private) Vector	simObjects Objects to be fired in the input sequence.

Fields inherited from class neuroidnet.phasesegregator.Peripheral

eventIterator, events, inputAreas, nextTime, percepts, segregation, sensoryArea

Fields inherited from class neuroidnet.ntr.Peripheral

network, time

Constructor Summary

StatisticalInputSequence(Network network)
In order to be used from bsh: As constructor only take necessary parameters, like connect to a network.

Method Summary

void	build() Builds properties of the peripheral object.
SortedSet	correctConcepts() Builds a sorted set from simObjects.
void	evaluateResults() After simulation, checks if desired concepts were allocated in the NTR during simulation.
String	getProperties() In addition to getStatus() contents, give all static properties also.
(package private) void	initEvents() Sets the eventIterator to first position.
double	qualityOfConcepts(Set concepts) Calculates a quality value in the range [0, 1], 1 indicating full replication has been reached in recruitment for each intended simulation object.
SortedSet	spuriousConcepts() Computes the set of spurious concepts by pruning from the contents of the ConceptArea the subsets of all simObjects.

Methods inherited from class neuroidnet.phasesegregator.Peripheral

createSensoryInputs, eventsAtThisTime, fireObjectInArea, getNext, testOneInput

Methods inherited from class neuroidnet.ntr.Peripheral

getStatus, step, toString

Methods inherited from class java.lang.Object

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

Field Detail

numberOfObjects

int numberOfObjects

Can be manipulated by beanshell scripts under scripts/.

numberOfNeuroids

int numberOfNeuroids

replication

int replication

numberOfMedialAreas

int numberOfMedialAreas

numberOfItemsPerArea

int numberOfItemsPerArea

simObjects

Vector simObjects

Objects to be fired in the input sequence.

Constructor Detail

StatisticalInputSequence

public StatisticalInputSequence(Network network)

In order to be used from bsh: As constructor only take necessary parameters, like connect to a network. Then set properties from bsh, and then call build to validate the object, which should call initEvents.. Needs changing previous architecture.

Parameters:

network - a Network value

Method Detail

build

public void build()

Builds properties of the peripheral object. Here, prepare random objects to be registered.

initEvents

void initEvents()

Description copied from class: Peripheral

Sets the eventIterator to first position.

Overrides:

initEvents in class Peripheral

See Also:

Peripheral.eventIterator, Peripheral.getNext()

evaluateResults

public void evaluateResults()

After simulation, checks if desired concepts were allocated in the NTR during simulation. For all objects, check if concepts has been formed and how many members do they have. Sets SimObject.concept

Simplification: Check only final concepts (anyway if intermediate concepts are not properly defined, final concepts cannot be formed.)

TODO: Produce a graph?

qualityOfConcepts

public double qualityOfConcepts(Set concepts)

Calculates a quality value in the range [0, 1], 1 indicating full replication has been reached in recruitment for each intended simulation object.

Parameters:

concepts - Concept set for which the quality is to be calculated.

Returns:

1 if all concepts contained replication elements, 0 if any of the concepts were missing, and something in between otherwise.

spuriousConcepts

public SortedSet spuriousConcepts()

Computes the set of spurious concepts by pruning from the contents of the ConceptArea the subsets of all simObjects.

Returns:

Set of all spurious concepts.

correctConcepts

public SortedSet correctConcepts()

Builds a sorted set from simObjects.

Returns:

Set of recruited correct concepts.

getProperties

public String getProperties()

Description copied from interface: Expressive

In addition to getStatus() contents, give all static properties also.

Specified by:

getProperties in interface Expressive

Overrides:

getProperties in class Peripheral

Returns: