edu.stanford.nlp.linalg
Class EAlgebra

java.lang.Object
  |
  +--cern.colt.PersistentObject
        |
        +--cern.colt.matrix.linalg.Algebra
              |
              +--edu.stanford.nlp.linalg.EAlgebra

All Implemented Interfaces:

Cloneable, Serializable

public class EAlgebra

extends cern.colt.matrix.linalg.Algebra

Extends cern.colt.matrix.linalg.Algebra to add some functions that are useful in NLP.

See Also:

Serialized Form

Field Summary

Fields inherited from class cern.colt.matrix.linalg.Algebra

DEFAULT, property, ZERO

Fields inherited from class cern.colt.PersistentObject

serialVersionUID

Constructor Summary

EAlgebra()

Method Summary

cern.colt.matrix.DoubleMatrix2D	center(cern.colt.matrix.DoubleMatrix2D A) Centers the matrix around its mean.
cern.colt.matrix.DoubleMatrix1D	mean(cern.colt.matrix.DoubleMatrix2D A) evaluates mean of columns in A
cern.colt.matrix.DoubleMatrix1D	principalDirection(cern.colt.matrix.DoubleMatrix2D A) calculates the first eigenvector of the autocorrelation matrix A*Atranspose (the first singular vector of the matrix A) right now the implementation is very slow.

Methods inherited from class cern.colt.matrix.linalg.Algebra

clone, cond, det, hypot, hypotFunction, inverse, mult, mult, mult, multOuter, norm1, norm1, norm2, norm2, normF, normInfinity, normInfinity, permute, permute, permuteColumns, permuteRows, pow, property, rank, setProperty, solve, solveTranspose, subMatrix, toString, toVerboseString, trace, transpose, trapezoidalLower

Methods inherited from class java.lang.Object

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

Constructor Detail

EAlgebra

public EAlgebra()

Method Detail

mean

public cern.colt.matrix.DoubleMatrix1D mean(cern.colt.matrix.DoubleMatrix2D A)

evaluates mean of columns in A

center

public cern.colt.matrix.DoubleMatrix2D center(cern.colt.matrix.DoubleMatrix2D A)

Centers the matrix around its mean. That is, it subtracts off the mean from each column of the matrix.

principalDirection

public cern.colt.matrix.DoubleMatrix1D principalDirection(cern.colt.matrix.DoubleMatrix2D A)

calculates the first eigenvector of the autocorrelation matrix A*Atranspose (the first singular vector of the matrix A) right now the implementation is very slow. calculates the entire SVD