edu.ucla.sspace.matrix.factorization

Class AbstractSvd

java.lang.Object

edu.ucla.sspace.matrix.factorization.AbstractSvd

All Implemented Interfaces:

MatrixFactorization

Direct Known Subclasses:

SingularValueDecompositionLibJ, SingularValueDecompositionMatlab, SingularValueDecompositionOctave

public abstract class AbstractSvd
extends Object
implements MatrixFactorization

An abstract base class for any Singular Value Decomposition implementation. This base class simply provides protected Matrix data members corresponding to the U,S,and V matrices resulting from the SVD. It also automatically computes U*S when calling dataClasses and S*Vt when calling classFeatures, both of which are done lazily and then cached so that the multiplication is done only once.

Subclasses need only implement MatrixFactorization.factorize(MatrixFile, int) and factorization(SparseMatrix, int), and MatrixFactorization.getBuilder().

Author:

Keith Stevens

Field Summary

Fields

Modifier and Type	Field and Description
protected Matrix	classFeatures The class by feature type matrix.
protected Matrix	dataClasses The data point by class matrix.
protected boolean	scaledClassFeatures Set to true when classFeatures has been accessed the first time to mark that the singular values have been applied to each value in the matrix.
protected boolean	scaledDataClasses Set to true when dataClasses has been accessed the first time to mark that the singular values have been applied to each value in the matrix.
protected double[]	singularValues The singular values computed during factorization.

Constructor Summary

Constructors

Constructor and Description
AbstractSvd()

Method Summary

Methods

Modifier and Type	Method and Description
Matrix	classFeatures() Returns the latent class by feature Matrix.
Matrix	dataClasses() Returns the data point by latent class Matrix.
double[]	singularValues() Returns a double array of the singular values.

Methods inherited from class java.lang.Object

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

Methods inherited from interface edu.ucla.sspace.matrix.MatrixFactorization

factorize, factorize, getBuilder

Field Detail

classFeatures

protected Matrix classFeatures

The class by feature type matrix.

scaledClassFeatures

protected boolean scaledClassFeatures

Set to true when classFeatures has been accessed the first time to mark that the singular values have been applied to each value in the matrix.

dataClasses

protected Matrix dataClasses

The data point by class matrix.

scaledDataClasses

protected boolean scaledDataClasses

Set to true when dataClasses has been accessed the first time to mark that the singular values have been applied to each value in the matrix.

singularValues

protected double[] singularValues

The singular values computed during factorization.

Constructor Detail

AbstractSvd

public AbstractSvd()

Method Detail

dataClasses

public Matrix dataClasses()

Returns the data point by latent class Matrix. This matrix represents the degree by which each data point can be explained by the discovered latent classes. Actual interpretations of the latent classes depends on the actual algorithm used. This Matrix can be used as a reduced representation of the data points themselves.

Specified by:

dataClasses in interface MatrixFactorization

classFeatures

public Matrix classFeatures()

Returns the latent class by feature Matrix. This matrix represents the degree by which each feature affects each latent class. Actual interpretations of this interaction depends on the actual algorithm used. This Matrix can be used as a reduced representation of the features.

Specified by:

classFeatures in interface MatrixFactorization

singularValues

public double[] singularValues()

Returns a double array of the singular values.