uchicago.src.repastdemos.neural
Class Employee

java.lang.Object
  uchicago.src.sim.network.DefaultNode
      uchicago.src.sim.network.DefaultDrawableNode
          uchicago.src.repastdemos.neural.Employee

All Implemented Interfaces:

AutoStepable, Drawable2DGridNode, DrawableNonGridNode, Moveable, NonGridDrawable, Node

Direct Known Subclasses:

TrainedEmployee

public class Employee
extends DefaultDrawableNode
implements AutoStepable

This class represents the agents who are to perform an action for the Boss. The agent learns to perform the correct action by training a neural network based on the commands received, and whether they were scolded for the action they did perform, or praised.

Version:

$Revision: 1.1 $ $Date: 2005/08/12 20:04:54 $

Author:

Jerry Vos

Nested Class Summary

Nested classes/interfaces inherited from class uchicago.src.sim.network.DefaultNode

DefaultNode.OrderedHashMap

Field Summary

static int	BOSS_A
static int	BOSS_B
protected org.joone.io.MemoryInputSynapse	desiredNetworkOutput This is used to store the output that we want to train the network towards
static int	DO_NOTHING
static int	DO_SOMETHING
protected org.joone.engine.DirectSynapse	inputForRetrieval This is the input synapse used to grab information from the network.
protected org.joone.io.MemoryInputSynapse	inputForTraining This is used to store the input that results in the desired output
protected RepastNeuralWrapper	net This is the neural network used by this employee

Fields inherited from class uchicago.src.sim.network.DefaultDrawableNode

item

Fields inherited from class uchicago.src.sim.network.DefaultNode

inEdges, inMap, label, outEdges, outMap

Constructor Summary

Employee()

Employee(double x, double y)

Method Summary

protected void	doNothing()
protected void	doSomething()
void	draw(SimGraphics g) Draws this Drawable2DGridNode.
double	getActionPerformed()
double[]	getCommands()
double	getError()
void	postStep() normally the method for swapping in new public data and otherwise handling the results/cleanup of a step
void	praise(Boss boss)
void	preStep() normally the method for setting up an agent for the next step
void	receiveCommand(int bossID, int command)
static void	resetIndices()
void	scold(Boss boss)
void	step() normally the method that performs the actual actions of a step

Methods inherited from class uchicago.src.sim.network.DefaultDrawableNode

allowResizing, calcSize, contains, getBorderColor, getBorderWidth, getColor, getFont, getHeight, getLabelColor, getNode, getWidth, getX, getY, setBorderColor, setBorderWidth, setColor, setDrawable, setDrawableNoCopy, setFont, setHeight, setLabelColor, setNodeLabel, setWidth, setX, setX, setY, setY

Methods inherited from class uchicago.src.sim.network.DefaultNode

addInEdge, addInEdges, addOutEdge, addOutEdges, clearInEdges, clearOutEdges, getEdgesFrom, getEdgesTo, getFromNodes, getId, getInDegree, getInEdges, getInNodes, getNodeLabel, getNumInEdges, getNumOutEdges, getOutDegree, getOutEdges, getOutNodes, getRandomFromNode, getRandomNodeIn, getRandomNodeOut, getRandomToNode, getToNodes, hasEdgeFrom, hasEdgeTo, hasEdgeToOrFrom, makeRandomInEdge, makeRandomOutEdge, removeEdgesFrom, removeEdgesTo, removeInEdge, removeOutEdge

Methods inherited from class java.lang.Object

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

Methods inherited from interface uchicago.src.sim.gui.DrawableNonGridNode

getOutEdges

Methods inherited from interface uchicago.src.sim.gui.Drawable2DGridNode

getOutEdges

Field Detail

DO_NOTHING

public static final int DO_NOTHING

See Also:

Constant Field Values

DO_SOMETHING

public static final int DO_SOMETHING

See Also:

Constant Field Values

BOSS_A

public static final int BOSS_A

See Also:

Constant Field Values

BOSS_B

public static final int BOSS_B

See Also:

Constant Field Values

net

protected RepastNeuralWrapper net

This is the neural network used by this employee

desiredNetworkOutput

protected org.joone.io.MemoryInputSynapse desiredNetworkOutput

This is used to store the output that we want to train the network towards

inputForTraining

protected org.joone.io.MemoryInputSynapse inputForTraining

This is used to store the input that results in the desired output

inputForRetrieval

protected org.joone.engine.DirectSynapse inputForRetrieval

This is the input synapse used to grab information from the network. DirectSynapses work the best for this type of usage, that is, when a single input pattern will be applied to the network.

Constructor Detail

Employee

public Employee(double x,
                double y)
         throws RepastException

Throws:

RepastException

Employee

public Employee()
         throws RepastException

Throws:

RepastException

Method Detail

resetIndices

public static void resetIndices()

preStep

public void preStep()
             throws NeuralException

Description copied from interface: AutoStepable

normally the method for setting up an agent for the next step

Specified by:

preStep in interface AutoStepable

Throws:

NeuralException

step

public void step()
          throws NeuralException

Description copied from interface: AutoStepable

normally the method that performs the actual actions of a step

Specified by:

step in interface AutoStepable

Throws:

NeuralException

postStep

public void postStep()

Description copied from interface: AutoStepable

normally the method for swapping in new public data and otherwise handling the results/cleanup of a step

Specified by:

postStep in interface AutoStepable

draw

public void draw(SimGraphics g)

Description copied from interface: Drawable2DGridNode

Draws this Drawable2DGridNode.

Specified by:

draw in interface Drawable2DGridNode

Specified by:

draw in interface NonGridDrawable

Overrides:

draw in class DefaultDrawableNode

Parameters:

g - the SimGraphics object used for drawing

scold

public void scold(Boss boss)

praise

public void praise(Boss boss)

receiveCommand

public void receiveCommand(int bossID,
                           int command)

doNothing

protected void doNothing()

doSomething

protected void doSomething()

getActionPerformed

public double getActionPerformed()

getCommands

public double[] getCommands()

getError

public double getError()

Returns:

returns the neural network's error