Class Square

java.lang.Object
  |
  +--Square

All Implemented Interfaces:

Environment, java.io.Serializable

public class Square

extends java.lang.Object

implements Environment, java.io.Serializable

This class implements the problem to be solved by agents. It first generates randomly two values corresponding to the two parameters taken into account for the evaluation. Then, it calculates the value corresponding to the combination of these two variables. These two variables are then converted in a binary string that is the signal transmitted to the agent. It then calculates the reward, comparing the calculated output and the answer provided by the agent.

Since:

JDK 1.3

See Also:

Serialized Form

Field Summary

private int	bits the number of bits used to code the variables in binaries
private int[]	car The different variables are stored in an array of integers.
static int	conLength Specifies the length of each presented problem.
private static XCSConstants	cons Stores the relevant constants for XCS.
private boolean	correct Stores if the last classification was correct.
private java.lang.String	currentState Stores the current problem.
private int	edgeSquare Edge of the square problem
private int	maxPayoff Specifies the maximal payoff possible in this environment.
private int	nbVar The number of variables taken into account in the process of evaluation, i.e. in calculating the right action.
private int	nrActions In the square problem there are 6 possible classifications
static boolean	payoffLandscape Defines if either a payoff landscape or a 1000/0 payoff is provided after the execution of a classification.
private java.util.Random	random Used to generate new situations randomly
private boolean	reset Is set to true after a classification was executed
private int	scale In the square problem, variables'values can rank from 0 to 3

Constructor Summary

Square(int payoffMap)
Constructs the square problem according to the specified problem length and chosen payoff type.

Method Summary

double	computeDistance(java.lang.String string1, java.lang.String string2) Computes the distance between two points in the problem space.
boolean	doReset() Returns true after the current problem was classified
private java.lang.String	encode(int n) turn an integer into a binary
double	executeAction(int action) Executes the action and determines the reward.
private int[]	generateSituation() Generates the values of the variables constituting the current state
int	getConditionLength() Returns the problem length
java.lang.String	getCurrentState() Returns the current problem
int	getMaxPayoff() Returns the maximal payoff possible in the current multiplexer problem.
int	getNbVar() Returns the number of variables comprised in the problem
int	getNrActions() Returns the number of possible actions.
double	getProblemSurface() Returns the edge of the surface of the square problem space
boolean	isMultiStepProblem() Returns false since the square problem is a single step problem
java.lang.String	resetState() Generates a new random problem instance.
boolean	wasCorrect() Returns true if the last executed action was a correct classification

Methods inherited from class java.lang.Object

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

Field Detail

conLength

public static int conLength

Specifies the length of each presented problem.

payoffLandscape

public static boolean payoffLandscape

Defines if either a payoff landscape or a 1000/0 payoff is provided after the execution of a classification.

maxPayoff

private int maxPayoff

Specifies the maximal payoff possible in this environment.

edgeSquare

private int edgeSquare

Edge of the square problem

random

private java.util.Random random

Used to generate new situations randomly

currentState

private java.lang.String currentState

Stores the current problem.

cons

private static XCSConstants cons

Stores the relevant constants for XCS.

See Also:

XCSConstants

correct

private boolean correct

Stores if the last classification was correct.

reset

private boolean reset

Is set to true after a classification was executed

nrActions

private final int nrActions

In the square problem there are 6 possible classifications

See Also:

Constant Field Values

scale

private final int scale

In the square problem, variables'values can rank from 0 to 3

See Also:

Constant Field Values

bits

private int bits

the number of bits used to code the variables in binaries

nbVar

private final int nbVar

The number of variables taken into account in the process of evaluation, i.e. in calculating the right action.

See Also:

Constant Field Values

car

private int[] car

The different variables are stored in an array of integers. This array is used to generate the string passed to the agents and the value of the correct action

Constructor Detail

Square

public Square(int payoffMap)

Constructs the square problem according to the specified problem length and chosen payoff type. Essentially the relevant constants for the environment are calculated here. Moreover, the problem array is generated

Parameters:

payoffMap - Specifies if a payoff map should be provided or a 1000/0 payoff.

Method Detail

resetState

public java.lang.String resetState()

Generates a new random problem instance.

Specified by:

resetState in interface Environment

See Also:

generateSituation(), encode(int)

executeAction

public double executeAction(int action)

Executes the action and determines the reward. Distinguishes between the payoff landscape and the 0/1000 reward.

Specified by:

executeAction in interface Environment

Parameters:

action - Specifies the classification.

generateSituation

private int[] generateSituation()

Generates the values of the variables constituting the current state

encode

private java.lang.String encode(int n)

turn an integer into a binary

Parameters:

n - The integer to be turned into a binary string

wasCorrect

public boolean wasCorrect()

Returns true if the last executed action was a correct classification

Specified by:

wasCorrect in interface Environment

isMultiStepProblem

public boolean isMultiStepProblem()

Returns false since the square problem is a single step problem

Specified by:

isMultiStepProblem in interface Environment

doReset

public boolean doReset()

Returns true after the current problem was classified

Specified by:

doReset in interface Environment

getConditionLength

public int getConditionLength()

Returns the problem length

Specified by:

getConditionLength in interface Environment

getMaxPayoff

public int getMaxPayoff()

Returns the maximal payoff possible in the current multiplexer problem. The maximal payoff is determined out of the payoff type. If the payoff type 1000/0 is selected, this function will return 1000, otherwise the maximal value depends on the problem size.

Specified by:

getMaxPayoff in interface Environment

getNrActions

public int getNrActions()

Returns the number of possible actions. in the square problem there are six classifications possible

Specified by:

getNrActions in interface Environment

getNbVar

public int getNbVar()

Returns the number of variables comprised in the problem

getCurrentState

public java.lang.String getCurrentState()

Returns the current problem

Specified by:

getCurrentState in interface Environment

computeDistance

public double computeDistance(java.lang.String string1,
                              java.lang.String string2)

Computes the distance between two points in the problem space. The coordinates are written as binary strings. This method is used to compute the distance between the signal from the environment and the classifier most remote from it, in order to assess whether or not the signal is in the realm of competence of the agent (or XCS) to whom the signal is submitted.

Specified by:

computeDistance in interface Environment

Parameters:

string1 - Coordinates of the first position used in the comparison

string2 - Coordinates of the second position used in the comparison

getProblemSurface

public double getProblemSurface()

Returns the edge of the surface of the square problem space

Specified by:

getProblemSurface in interface Environment