JSci.maths.wavelet

Class MatchingPursuit

java.lang.Object

JSci.maths.wavelet.BasisFunctionLibrary

JSci.maths.wavelet.MatchingPursuit

All Implemented Interfaces:

java.lang.Cloneable

public class MatchingPursuit
extends BasisFunctionLibrary
implements java.lang.Cloneable

A Wavelet (and more) matching pursuit class Uses adaptative Morse coding for better performance. The MatchingPursuit is used to obtain a Time-Frequency representation (TF) through a fast algorithm.

Author:

Daniel Lemire

Field Summary

Fields inherited from class JSci.maths.wavelet.BasisFunctionLibrary

DFunction, Fdual, Fprimary

Constructor Summary

Constructors

Constructor and Description
MatchingPursuit(DiscreteFunction f) Constructor

Method Summary

Methods

Modifier and Type	Method and Description
void	add(MultiscaleFunction fprimary, MultiscaleFunction fdual) Add a clone of the given MultiscaleFunctions to the internal arrays of MultiscaleFunctions.
java.lang.Object	clone()
void	diagnostic(double tol) Check the validity of the current matching algorithm.
double[][]	forcedMatch(int pos) Force the system to select the given element as the best match.
double[]	getCoefs() Recover the matching coefficients.
int[]	getRecord() all matches are recorded so one can trace them back
double[]	getRecordedNorms() Trace back how much of the norm was taken out at each match.
double[][]	match() Does the matching 1 time and return the TF representation.
double[][]	match(int j) Does the matching j times and return the TF representation.
double[][]	matchAll(double tol) Repeatly match until it remains less than tol * 100 percent of the original L2 norm; no matter what, at least one match will be done.
void	setData(DiscreteFunction f) Allows the user to change the DiscreteFunction (see constructor).

Methods inherited from class JSci.maths.wavelet.BasisFunctionLibrary

add, add, add, add, checkBiorthogonality, cloneArrayDiscreteFunction, getData, getDual, getMorseThreshold, getPrimary, getResidue, getResidues, getSize, getWeigth, getWeigths, includeFourier, includeMasslessFourier, norm, norm, setMorseThreshold

Methods inherited from class java.lang.Object

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

Constructor Detail

MatchingPursuit

public MatchingPursuit(DiscreteFunction f)

Constructor

Method Detail

add

public void add(MultiscaleFunction fprimary,
       MultiscaleFunction fdual)

Description copied from class: BasisFunctionLibrary

Add a clone of the given MultiscaleFunctions to the internal arrays of MultiscaleFunctions.

Overrides:

add in class BasisFunctionLibrary

clone

public java.lang.Object clone()

Overrides:

clone in class BasisFunctionLibrary

getRecord

public int[] getRecord()

all matches are recorded so one can trace them back

getRecordedNorms

public double[] getRecordedNorms()

Trace back how much of the norm was taken out at each match.

getCoefs

public double[] getCoefs()

Recover the matching coefficients.

setData

public void setData(DiscreteFunction f)

Description copied from class: BasisFunctionLibrary

Allows the user to change the DiscreteFunction (see constructor).

Overrides:

setData in class BasisFunctionLibrary

diagnostic

public void diagnostic(double tol)
                throws MaximumIterationsExceededException

Check the validity of the current matching algorithm. Will return an exception if the diagnostic fails. This makes sense: the software should stop if the algorithm isn't safely within the given tolerance.

Throws:

java.lang.IllegalArgumentException - if the matching fails

MaximumIterationsExceededException - if it can't match one of the elements of the dictionnary

java.lang.IllegalArgumentException - if tol is is negative

match

public double[][] match()

Does the matching 1 time and return the TF representation. The TF representation may contain negative values and so, taking the absolute value of the result will often be useful.

matchAll

public double[][] matchAll(double tol)
                    throws MaximumIterationsExceededException

Repeatly match until it remains less than tol * 100 percent of the original L2 norm; no matter what, at least one match will be done. It will return the TF representaiton. The TF representation may contain negative values and so, taking the absolute value of the result will often be useful.

Parameters:

tol - percentile of energy

Throws:

java.lang.IllegalArgumentException - if tol is not within the interval [0,1]

MaximumIterationsExceededException - if the number of required match exceeds 5 times the size of the dictionnary (it should be a more than confortable margin unless the problem is ill-posed, change the dictionnary if it doesn't work)

forcedMatch

public double[][] forcedMatch(int pos)

Force the system to select the given element as the best match. It will return the TF representaiton. The TF representation may contain negative values and so, taking the absolute value of the result will often be useful.

match

public double[][] match(int j)

Does the matching j times and return the TF representation. The TF representation may contain negative values and so, taking the absolute value of the result will often be useful.

Parameters:

number - of iterations

Throws:

java.lang.IllegalArgumentException - if j is not positive