org.kramerlab.bmad.matrix

Class BooleanMatrix

java.lang.Object

org.kramerlab.bmad.matrix.BooleanMatrix

public class BooleanMatrix
extends Object

Dense boolean matrix with 3-valued boolean logic TRUE is represented as 2, UNKNOWN is represented as 1, FALSE is represented as 0, AND corresponds to MIN, OR corresponds to MAX

Field Summary

Fields

Modifier and Type	Field and Description
static byte	FALSE
static byte	TRUE
static byte	UNKNOWN

Constructor Summary

Constructors

Constructor and Description
BooleanMatrix(BooleanMatrix b) Creates a deep copy of another matrix
BooleanMatrix(byte[][] rowMajor) Constructs dense boolean matrix from given array in row major format.
BooleanMatrix(weka.core.Instances instances) Constructs dense boolean matrix from weka instances, that is expected to be filled with values 0, ?, 1
BooleanMatrix(int h, int w) Constructs new empty matrix with given dimensions, filled with FALSE

Method Summary

Methods

Modifier and Type	Method and Description
byte	apply(int c)
byte	apply(int r, int c)
void	baxoy(byte alpha, BooleanMatrix x) In-place operation on two row vectors, corresponding to SAXPY in linear algebra
BooleanMatrix	booleanProduct(BooleanMatrix other) Calculates the boolean product with the other matrix.
int[]	elementCount() Counts ones, unknowns, and zeros contained in this matrix
int	getHeight()
BooleanMatrix	getRow(int r) Extracts single row as separate matrix.
BooleanMatrix	getRows(List<Integer> indices)
int	getWidth()
BooleanMatrix	mapBoolean(Function<Byte,Byte> function) Map-method, restricted to boolean matrices as output.
static byte	not(byte b)
double	reconstructionError(BooleanMatrix reconstruction, double onesWeight) Calculates the reconstruction error between this matrix and the reconstruction.
Tuple<Double,Double>	relativeOneZeroZeroOneReconstructionError(BooleanMatrix reconstruction, double onesWeight) Calculates the relative 1->0 and 0->1 reconstruction error, that is, the total error divided by total maximum possible error.
double	relativeReconstructionError(BooleanMatrix reconstruction, double onesWeight) Calculates the relative error (error divided by the maximum possible error).
void	setRow(int r, BooleanMatrix row) Analogous to get row: no values are copied
Tuple<Integer,Integer>	size() Returns width and height of the matrix.
Image	toImage() Draws this matrix as image.
weka.core.Instances	toInstances() Converts the matrix to Weka-Instances
RowMajor<Byte>	toRowMajor() Transforms this matrix to a generic matrix.
String	toString()
void	update(int c, byte b)
void	update(int r, int c, byte b) Sets the entry at position (r, c) to b

Methods inherited from class java.lang.Object

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

Field Detail

TRUE

public static final byte TRUE

See Also:

Constant Field Values

UNKNOWN

public static final byte UNKNOWN

See Also:

Constant Field Values

FALSE

public static final byte FALSE

See Also:

Constant Field Values

Constructor Detail

BooleanMatrix

public BooleanMatrix(int h,
             int w)

Constructs new empty matrix with given dimensions, filled with FALSE

Parameters:

h - height

w - width

BooleanMatrix

public BooleanMatrix(BooleanMatrix b)

Creates a deep copy of another matrix

Parameters:

b -

BooleanMatrix

public BooleanMatrix(byte[][] rowMajor)

Constructs dense boolean matrix from given array in row major format. Assumes that the array is rectangular (all rows have equal length), and not degenerate (has entries).

Parameters:

rowMajor - entries in column major format

BooleanMatrix

public BooleanMatrix(weka.core.Instances instances)

Constructs dense boolean matrix from weka instances, that is expected to be filled with values 0, ?, 1

Method Detail

not

public static byte not(byte b)

getWidth

public int getWidth()

getHeight

public int getHeight()

update

public void update(int r,
          int c,
          byte b)

Sets the entry at position (r, c) to b

Parameters:

r - row

c - column

b - 0, 1 or 2

apply

public byte apply(int r,
         int c)

apply

public byte apply(int c)

update

public void update(int c,
          byte b)

getRows

public BooleanMatrix getRows(List<Integer> indices)

baxoy

public void baxoy(byte alpha,
         BooleanMatrix x)

In-place operation on two row vectors, corresponding to SAXPY in linear algebra

Parameters:

alpha - factor

x - other row

getRow

public BooleanMatrix getRow(int r)

Extracts single row as separate matrix. No values are copied, the row is backed by the same array

Parameters:

r - index of the row

Returns:

row matrix

setRow

public void setRow(int r,
          BooleanMatrix row)

Analogous to get row: no values are copied

Parameters:

r - index of the row

row - row vector

booleanProduct

public BooleanMatrix booleanProduct(BooleanMatrix other)

Calculates the boolean product with the other matrix.

Parameters:

other - boolean matrix with compatible dimension

Returns:

toRowMajor

public RowMajor<Byte> toRowMajor()

Transforms this matrix to a generic matrix. The generic versions are significantly slower for computations, but provide many handy O(n^2) methods (e.g. for printing and drawing matrices)

Returns:

same matrix in generic format

reconstructionError

public double reconstructionError(BooleanMatrix reconstruction,
                         double onesWeight)

Calculates the reconstruction error between this matrix and the reconstruction.

Parameters:

reconstruction - some other matrix

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

Returns:

relativeOneZeroZeroOneReconstructionError

public Tuple<Double,Double> relativeOneZeroZeroOneReconstructionError(BooleanMatrix reconstruction,
                                                             double onesWeight)

Calculates the relative 1->0 and 0->1 reconstruction error, that is, the total error divided by total maximum possible error. Assumes, that there are no unknowns in the reconstruction. The total weight equals #0 + onesWeight * #1. Each 1->0 error costs onesWeight. Each 0->1 error costs 1.

Parameters:

reconstruction -

onesWeight -

Returns:

tuple with relative 1->0 and 0->1 errors

relativeReconstructionError

public double relativeReconstructionError(BooleanMatrix reconstruction,
                                 double onesWeight)

Calculates the relative error (error divided by the maximum possible error).

Parameters:

reconstruction -

onesWeight - relative weight of 1->0 errors

Returns:

total relative error

toString

public String toString()

Overrides:

toString in class Object

toImage

public Image toImage()

Draws this matrix as image. False is white, true is black, unknown is gray.

Returns:

image representation of the matrix.

toInstances

public weka.core.Instances toInstances()

Converts the matrix to Weka-Instances

Returns:

instances

mapBoolean

public BooleanMatrix mapBoolean(Function<Byte,Byte> function)

Map-method, restricted to boolean matrices as output. Creates a new matrix, that consists of point-wise applications of the function on the entries of this matrix.

Parameters:

function -

Returns:

size

public Tuple<Integer,Integer> size()

Returns width and height of the matrix.

Returns:

elementCount

public int[] elementCount()

Counts ones, unknowns, and zeros contained in this matrix

Returns:

array with number of zeros, unknowns and ones (in this order)