org.kramerlab.bmad.algorithms

Class GreedySelector

java.lang.Object

org.kramerlab.bmad.algorithms.GreedySelector

All Implemented Interfaces:

BasisSelector

public class GreedySelector
extends Object
implements BasisSelector

Basis Selector algorithm, that greedily selects the basis rows, estimating the reconstruction error by keeping an approximation of the combination matrix, which is also updated greedily in each step. Similar to what was proposed as second step of Miettinen's DBP algorithm.

Constructor Summary

Constructors

Constructor and Description
GreedySelector()

Method Summary

Methods

Modifier and Type	Method and Description
Tuple<BooleanMatrix,BooleanMatrix>	selectBasis(BooleanMatrix candidates, BooleanMatrix a, int dimension, double onesWeight) Calculates a boolean basis matrix B (second factor), and (optionally) a coarse approximation of the combination matrix C (first factor).
String	toString()

Methods inherited from class java.lang.Object

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

Constructor Detail

GreedySelector

public GreedySelector()

Method Detail

selectBasis

public Tuple<BooleanMatrix,BooleanMatrix> selectBasis(BooleanMatrix candidates,
                                             BooleanMatrix a,
                                             int dimension,
                                             double onesWeight)

Description copied from interface: BasisSelector

Calculates a boolean basis matrix B (second factor), and (optionally) a coarse approximation of the combination matrix C (first factor).

Specified by:

selectBasis in interface BasisSelector

Parameters:

candidates - matrix that contains candidate patterns in it's rows

a - original matrix

dimension - number of candidate rows to be included into the basis matrix

onesWeight - relative weight of 1 -> 0 errors, opposed to 0 -> 1 errors

Returns:

(combination, basis)

toString

public String toString()

Overrides:

toString in class Object