burlap.behavior.stochasticgames.agents.twoplayer.singlestage.equilibriumplayer

Class BimatrixEquilibriumSolver

java.lang.Object

burlap.behavior.stochasticgames.agents.twoplayer.singlestage.equilibriumplayer.BimatrixEquilibriumSolver

Direct Known Subclasses:

CorrelatedEquilibrium, MaxMax, MinMax, Utilitarian

public abstract class BimatrixEquilibriumSolver
extends java.lang.Object

This abstract class is used for computing the strategies according to a solution concept for a single stage bimatrix game. The solve(double[][], double[][]) method takes as input the payoff matrice for the row and column players of the Bimatrix game. If the Bimatrix is identical the to Bimatrix requested in the previous call to the solve method (if it was previously called), then nothing is done as the previous results are cached and can be retrieved. If the bimatrix is different, then the strategy for the row player and column player are computed using the abstract computeRowStrategy(double[][], double[][]) and computeColStrategy(double[][], double[][]) methods and then cached. The cahced results can be retreived using the getLastComputedRowStrategy() and getLastComputedColStrategy() methods.

Note that the bimatrix is compared to the last cached version for equality with an epislon difference comparions of the individual double values. That is, if the difference absolute difference |a - b| < doubleEpislon, then the a and b are treated as equal. By default, doubleEpislon is set to 1e-8; but this is a public static member that can be changed.

Author:

James MacGlashan

Field Summary

Fields

Modifier and Type	Field and Description
static double	doubleEpislon The epislon difference used to test for double equality.
protected double[][]	lastColPlayerPayoff The last cached column player payoff matrix
protected double[]	lastColsStrategy The last cached column player strategy
protected double[][]	lastRowPlayerPayoff The last cached row player payoff matrix
protected double[]	lastRowStrategy The last cached row player strategy

Constructor Summary

Constructors

Constructor and Description
BimatrixEquilibriumSolver()

Method Summary

Modifier and Type	Method and Description
protected boolean	bimatrixEqualsLast(double[][] rowPayoff, double[][] colPayoff) Tests whether the inputed bimatrix is equal to the last bimatrix cached by the solve(double[][], double[][]) method.
abstract double[]	computeColStrategy(double[][] rowPayoff, double[][] colPayoff) Computes and returns the column player strategy for the given bimatrix game.
abstract double[]	computeRowStrategy(double[][] rowPayoff, double[][] colPayoff) Computes and returns the row player strategy for the given bimatrix game.
protected static boolean	doubleEquality(double a, double b) Returns true if |a - b| < doubleEpislon; false otherwise.
double[]	getLastComputedColStrategy() Returns the last column player strategy computed by the solve(double[][], double[][]) method.
double[]	getLastComputedRowStrategy() Returns the last row player strategy computed by the solve(double[][], double[][]) method.
void	solve(double[][] rowPayoff, double[][] colPayoff) Solves and caches the solution for the given bimatrix.

Methods inherited from class java.lang.Object

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

Field Detail

doubleEpislon

public static final double doubleEpislon

The epislon difference used to test for double equality.

See Also:

Constant Field Values

lastRowStrategy

protected double[] lastRowStrategy

The last cached row player strategy

lastColsStrategy

protected double[] lastColsStrategy

The last cached column player strategy

lastRowPlayerPayoff

protected double[][] lastRowPlayerPayoff

The last cached row player payoff matrix

lastColPlayerPayoff

protected double[][] lastColPlayerPayoff

The last cached column player payoff matrix

Constructor Detail

BimatrixEquilibriumSolver

public BimatrixEquilibriumSolver()

Method Detail

solve

public void solve(double[][] rowPayoff,
                  double[][] colPayoff)

Solves and caches the solution for the given bimatrix. If the given bimatrix is the same as the last bimatrix solved by this object (if there was one), then nothing is computed and the same solution is returned.

Parameters:

rowPayoff - the row player payoff matrix

colPayoff - the column player payoff matrix

getLastComputedRowStrategy

public double[] getLastComputedRowStrategy()

Returns the last row player strategy computed by the solve(double[][], double[][]) method. The strategy is represented as a double array where a[i] is the probability of action i being selected.

Returns:

the last row player strategy.

getLastComputedColStrategy

public double[] getLastComputedColStrategy()

Returns the last column player strategy computed by the solve(double[][], double[][]) method. The strategy is represented as a double array where a[i] is the probability of action i being selected.

Returns:

the last column player strategy.

computeRowStrategy

public abstract double[] computeRowStrategy(double[][] rowPayoff,
                                            double[][] colPayoff)

Computes and returns the row player strategy for the given bimatrix game. The strategy is represented as a double array where a[i] is the probability of action i being selected.

Parameters:

rowPayoff - the row player payoffs.

colPayoff - the column player payoffs.

Returns:

the row player strategy.

computeColStrategy

public abstract double[] computeColStrategy(double[][] rowPayoff,
                                            double[][] colPayoff)

Computes and returns the column player strategy for the given bimatrix game. The strategy is represented as a double array where a[i] is the probability of action i being selected.

Parameters:

rowPayoff - the row player payoffs.

colPayoff - the column player payoffs.

Returns:

the column player strategy.

bimatrixEqualsLast

protected boolean bimatrixEqualsLast(double[][] rowPayoff,
                                     double[][] colPayoff)

Tests whether the inputed bimatrix is equal to the last bimatrix cached by the solve(double[][], double[][]) method. If there have been no previous calls to the solve(double[][], double[][]) method and therefore no cached bimatrix, then this method returns false. If the input bimatrix as a different dimensionality that the cached matrix, this method returns false. If any row or column player payoffs for the input bimatrix are different than cached bimatrix this method returns false. Otherwise this method returns true. Equality between two double values is tested with an epsilon difference where value a and b are evaluated as equal if |a - b| < doubleEpislon.

Parameters:

rowPayoff - the row player payoffs.

colPayoff - the column player payoffs.

Returns:

true if the input bimatrix is equal to the last cached bimatrix; false otherwise.

doubleEquality

protected static boolean doubleEquality(double a,
                                        double b)

Returns true if |a - b| < doubleEpislon; false otherwise.

Parameters:

a - first input

b - second input

Returns:

true if |a - b| < doubleEpislon; false otherwise.