de.lmu.ifi.dbs.elki.math.linearalgebra.randomprojections

Class AbstractRandomProjectionFamily.MatrixProjection

java.lang.Object

de.lmu.ifi.dbs.elki.math.linearalgebra.randomprojections.AbstractRandomProjectionFamily.MatrixProjection

All Implemented Interfaces:

RandomProjectionFamily.Projection

Enclosing class:

AbstractRandomProjectionFamily

public static class AbstractRandomProjectionFamily.MatrixProjection
extends Object
implements RandomProjectionFamily.Projection

Class to project using a matrix multiplication. This class is optimized for dense vector multiplications. In other words, the row dimensionality is the output dimensionality, the column dimensionality is the input dimensionality. It is not thread safe because it uses an internal buffer to store a local copy of the vector.

Author:

Erich Schubert

Field Summary

Fields

Modifier and Type	Field and Description
private double[]	buf Shared buffer to use during projections.
(package private) double[][]	matrix Projection matrix.

Constructor Summary

Constructors

Constructor and Description
AbstractRandomProjectionFamily.MatrixProjection(double[][] matrix) Constructor.

Method Summary

Methods

Modifier and Type	Method and Description
int	getOutputDimensionality() Get the output dimensionality.
double[]	project(NumberVector in) Project a single vector.
double[]	project(NumberVector in, double[] ret) Project a single vector, into the given buffer.
private double[]	projectSparse(SparseNumberVector in, double[] ret) Project, exploiting sparsity; but the transposed matrix layout would have been better.

Methods inherited from class java.lang.Object

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

Field Detail

matrix

double[][] matrix

Projection matrix.

buf

private double[] buf

Shared buffer to use during projections.

Constructor Detail

AbstractRandomProjectionFamily.MatrixProjection

public AbstractRandomProjectionFamily.MatrixProjection(double[][] matrix)

Constructor.

Parameters:

matrix - Projection matrix ([output dim][input dim]).

Method Detail

project

public double[] project(NumberVector in)

Description copied from interface: RandomProjectionFamily.Projection

Project a single vector.

Specified by:

project in interface RandomProjectionFamily.Projection

Parameters:

in - Input vector

Returns:

Projected vector

project

public double[] project(NumberVector in,
               double[] ret)

Description copied from interface: RandomProjectionFamily.Projection

Project a single vector, into the given buffer.

Specified by:

project in interface RandomProjectionFamily.Projection

Parameters:

in - Input vector

ret - Output buffer

Returns:

buffer

projectSparse

private double[] projectSparse(SparseNumberVector in,
                     double[] ret)

Project, exploiting sparsity; but the transposed matrix layout would have been better. For projections where you expect sparse input, consider the opposite.

Parameters:

in - Input vector

ret - Projection buffer

Returns:

Projected data.

getOutputDimensionality

public int getOutputDimensionality()

Description copied from interface: RandomProjectionFamily.Projection

Get the output dimensionality.

Specified by:

getOutputDimensionality in interface RandomProjectionFamily.Projection

Returns:

Output dimensionality