org.kramerlab.bmad.algorithms

Class GreedyCombinator

java.lang.Object

org.kramerlab.bmad.algorithms.Combinator

org.kramerlab.bmad.algorithms.GreedyCombinator

Direct Known Subclasses:

CoverGreedyCombinator, DensityGreedyCombinator

public abstract class GreedyCombinator
extends Combinator

Abstract greedy Combinator (third step of the algorithm). It greedily includes basis rows of maximal "usefulness", until no improvement is possible. Subclasses just have to override the usefulness method.

Constructor Summary

Constructors

Constructor and Description
GreedyCombinator()

Method Summary

Methods

Modifier and Type	Method and Description
BooleanMatrix	combineRow(BooleanMatrix row, BooleanMatrix _, BooleanMatrix basis, double onesWeight) Describes a row of the original matrix as combination of basis rows from the basis matrix.
protected abstract double	usefulness(Cover cover, BooleanMatrix row, BooleanMatrix basisRow)

Methods inherited from class org.kramerlab.bmad.algorithms.Combinator

combineMatrix

Methods inherited from class java.lang.Object

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

Constructor Detail

GreedyCombinator

public GreedyCombinator()

Method Detail

usefulness

protected abstract double usefulness(Cover cover,
                BooleanMatrix row,
                BooleanMatrix basisRow)

combineRow

public BooleanMatrix combineRow(BooleanMatrix row,
                       BooleanMatrix _,
                       BooleanMatrix basis,
                       double onesWeight)

Description copied from class: Combinator

Describes a row of the original matrix as combination of basis rows from the basis matrix.

Specified by:

combineRow in class Combinator

Parameters:

row - a row from the original matrix

_ - approximation, already obtained by previous steps

basis - the basis matrix (with basis patterns as rows)

onesWeight - relative weight of 1 -> 0 errors

Returns:

single row, that has a 0 or 1 entry for each basis row