org.apache.mahout.math

Class FileBasedSparseBinaryMatrix

java.lang.Object

org.apache.mahout.math.AbstractMatrix

org.apache.mahout.math.FileBasedSparseBinaryMatrix

All Implemented Interfaces:

Cloneable, Iterable<MatrixSlice>, Matrix, VectorIterable

public final class FileBasedSparseBinaryMatrix
extends AbstractMatrix

Provides a way to get data from a file and treat it as if it were a matrix, but avoids putting all that data onto the Java heap. Instead, the file is mapped into non-heap memory as a DoubleBuffer and we access that instead. The interesting aspect of this is that the values in the matrix are binary and sparse so we don't need to store the actual data, just the location of non-zero values.

Currently file data is formatted as follows:

• A magic number to indicate the file format.
• The size of the matrix (max rows and columns possible)
• Number of non-zeros in each row.
• A list of non-zero columns for each row. The list starts with a count and then has column numbers

It would be preferable to use something like protobufs to define the format so that we can use different row formats for different kinds of data. For instance, Golay coding of column numbers or compressed bit vectors might be good representations for some purposes.

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	FileBasedSparseBinaryMatrix.BinaryReadOnlyElement

Nested classes/interfaces inherited from class org.apache.mahout.math.AbstractMatrix

AbstractMatrix.TransposeViewVector

Field Summary

Fields inherited from class org.apache.mahout.math.AbstractMatrix

COL, columnLabelBindings, columns, ROW, rowLabelBindings, rows

Constructor Summary

Constructors

Constructor and Description
FileBasedSparseBinaryMatrix(int rows, int columns) Constructs an empty matrix of the given size.

Method Summary

Methods

Modifier and Type	Method and Description
Matrix	assignColumn(int column, Vector other) Assign the other vector values to the column of the receiver
Matrix	assignRow(int row, Vector other) Assign the other vector values to the row of the receiver
double	getQuick(int rowIndex, int columnIndex) Return the value at the given indexes, without checking bounds
Matrix	like() Return an empty matrix of the same underlying class as the receiver
Matrix	like(int rows, int columns) Returns an empty matrix of the same underlying class as the receiver and of the specified size.
void	setData(File f)
void	setQuick(int row, int column, double value) Set the value at the given index, without checking bounds
Matrix	viewPart(int[] offset, int[] size) Return a view into part of a matrix.
Vector	viewRow(int rowIndex) Returns a view of a row.
static void	writeMatrix(File f, Matrix m)

Methods inherited from class org.apache.mahout.math.AbstractMatrix

aggregate, aggregateColumns, aggregateRows, asFormatString, assign, assign, assign, assign, assign, clone, columnSize, determinant, divide, get, get, getColumnLabelBindings, getFlavor, getNumNondefaultElements, getRowLabelBindings, iterateAll, iterateNonEmpty, iterator, minus, numCols, numRows, numSlices, plus, plus, rowSize, set, set, set, set, set, set, setColumnLabelBindings, setRowLabelBindings, times, times, times, timesSquared, toString, transpose, viewColumn, viewDiagonal, viewPart, zSum

Methods inherited from class java.lang.Object

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

Constructor Detail

FileBasedSparseBinaryMatrix

public FileBasedSparseBinaryMatrix(int rows,
                           int columns)

Constructs an empty matrix of the given size.

Parameters:

rows - The number of rows in the result.

columns - The number of columns in the result.

Method Detail

setData

public void setData(File f)
             throws IOException

Throws:

IOException

writeMatrix

public static void writeMatrix(File f,
               Matrix m)
                        throws IOException

Throws:

IOException

assignColumn

public Matrix assignColumn(int column,
                  Vector other)

Assign the other vector values to the column of the receiver

Parameters:

column - the int row to assign

other - a Vector

Returns:

the modified receiver

Throws:

CardinalityException - if the cardinalities differ

assignRow

public Matrix assignRow(int row,
               Vector other)

Assign the other vector values to the row of the receiver

Parameters:

row - the int row to assign

other - a Vector

Returns:

the modified receiver

Throws:

CardinalityException - if the cardinalities differ

getQuick

public double getQuick(int rowIndex,
              int columnIndex)

Return the value at the given indexes, without checking bounds

Parameters:

rowIndex - an int row index

columnIndex - an int column index

Returns:

the double at the index

like

public Matrix like()

Return an empty matrix of the same underlying class as the receiver

Returns:

a Matrix

like

public Matrix like(int rows,
          int columns)

Returns an empty matrix of the same underlying class as the receiver and of the specified size.

Parameters:

rows - the int number of rows

columns - the int number of columns

setQuick

public void setQuick(int row,
            int column,
            double value)

Set the value at the given index, without checking bounds

Parameters:

row - an int row index into the receiver

column - an int column index into the receiver

value - a double value to set

viewPart

public Matrix viewPart(int[] offset,
              int[] size)

Return a view into part of a matrix. Changes to the view will change the original matrix.

Specified by:

viewPart in interface Matrix

Overrides:

viewPart in class AbstractMatrix

Parameters:

offset - an int[2] offset into the receiver

size - the int[2] size of the desired result

Returns:

a matrix that shares storage with part of the original matrix.

Throws:

CardinalityException - if the length is greater than the cardinality of the receiver

IndexException - if the offset is negative or the offset+length is outside of the receiver

viewRow

public Vector viewRow(int rowIndex)

Returns a view of a row. Changes to the view will affect the original.

Specified by:

viewRow in interface Matrix

Overrides:

viewRow in class AbstractMatrix

Parameters:

rowIndex - Which row to return.

Returns:

A vector that references the desired row.