org.kramerlab.bmad.algorithms

Class BooleanMatrixDecomposition

java.lang.Object

org.kramerlab.bmad.algorithms.BooleanMatrixDecomposition

public class BooleanMatrixDecomposition
extends Object

Implementation of a generic boolean matrix decomposition algorithm, that fits into this framework. Such an algorithm consists of three parts: candidate generator, basis selector, and combinator. This implementation basically invokes methods of the subcomponents in the correct order.

Field Summary

Fields

Modifier and Type	Field and Description
static BooleanMatrixDecomposition	BEST_UNCONFIGURED An algorithm that delivers best results (one the two examples we tested) without further configuration.
protected Combinator	combinator
static BooleanMatrixDecomposition	DBP Algorithm proposed by Miettinen et.
protected CandidateGenerator	generator
static BooleanMatrixDecomposition	LOC_ITER Algorithm proposed by Miettinen for the BCX problem.
protected double	onesWeight
protected BasisSelector	selector

Constructor Summary

Constructors

Constructor and Description
BooleanMatrixDecomposition(CandidateGenerator generator, BasisSelector selector, Combinator combinator, double onesWeight)

Method Summary

Methods

Modifier and Type	Method and Description
static BooleanMatrixDecomposition	BEST_CONFIGURED(double assocThreshold) Promising algorithm, that requires one further parameter to be learned.
Tuple<BooleanMatrix,BooleanMatrix>	decompose(BooleanMatrix a, int dimension) Decomposes matrix into two factor matrices.
Tuple<weka.core.Instances,weka.core.Instances>	decompose(weka.core.Instances a, int dimension) Thin wrapper around the other decompose method, that translates input and output into Weka Instances.
String	toString()

Methods inherited from class java.lang.Object

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

Field Detail

generator

protected CandidateGenerator generator

selector

protected BasisSelector selector

combinator

protected Combinator combinator

onesWeight

protected double onesWeight

LOC_ITER

public static BooleanMatrixDecomposition LOC_ITER

Algorithm proposed by Miettinen for the BCX problem.

DBP

public static BooleanMatrixDecomposition DBP

Algorithm proposed by Miettinen et. al for the DBP problem.

BEST_UNCONFIGURED

public static BooleanMatrixDecomposition BEST_UNCONFIGURED

An algorithm that delivers best results (one the two examples we tested) without further configuration.

Constructor Detail

BooleanMatrixDecomposition

public BooleanMatrixDecomposition(CandidateGenerator generator,
                          BasisSelector selector,
                          Combinator combinator,
                          double onesWeight)

Method Detail

decompose

public Tuple<BooleanMatrix,BooleanMatrix> decompose(BooleanMatrix a,
                                           int dimension)

Decomposes matrix into two factor matrices.

Parameters:

a - original matrix

dimension - number of basis rows

Returns:

(combinationMatrix, basisMatrix)

decompose

public Tuple<weka.core.Instances,weka.core.Instances> decompose(weka.core.Instances a,
                                                       int dimension)

Thin wrapper around the other decompose method, that translates input and output into Weka Instances.

Parameters:

a - original matrix

dimension - number of basis rows

Returns:

(combinationMatrix, basisMatrix)

toString

public String toString()

Overrides:

toString in class Object

BEST_CONFIGURED

public static BooleanMatrixDecomposition BEST_CONFIGURED(double assocThreshold)

Promising algorithm, that requires one further parameter to be learned.

Parameters:

assocThreshold - parameter for the candidate generation step

Returns:

Assoc+FastLoc+DensGreedy+Iter