Package edu.gsu.cs.dmlab.sparse.dictionary

Class CorrelationDictionaryCleaner

java.lang.Object

edu.gsu.cs.dmlab.sparse.dictionary.CorrelationDictionaryCleaner

All Implemented Interfaces:

IDictionaryCleaner

public class CorrelationDictionaryCleaner
extends Object
implements IDictionaryCleaner

Class used to update a dictionary used in sparse coding. This is generally done during the dictionary learning process. If elements are highly correlated one is selected for replacement with a randomly drawn signal from the input.

Author:

Dustin Kempton, Data Mining Lab, Georgia State University

Constructor Summary

Constructors

Constructor	Description
CorrelationDictionaryCleaner()

Method Summary

Modifier and Type	Method	Description
void	cleanDictionary(smile.math.matrix.DenseMatrix dictionary, List<double[]> input, smile.math.matrix.DenseMatrix gram)	Cleans the dictionary based on some algorithm given the input matrices.

Methods inherited from class java.lang.Object

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

Constructor Detail

CorrelationDictionaryCleaner

public CorrelationDictionaryCleaner()

Method Detail

cleanDictionary

public void cleanDictionary(smile.math.matrix.DenseMatrix dictionary,
                            List<double[]> input,
                            smile.math.matrix.DenseMatrix gram)

Description copied from interface: IDictionaryCleaner

Cleans the dictionary based on some algorithm given the input matrices.

Specified by:

cleanDictionary in interface IDictionaryCleaner

Parameters:

dictionary - Dictionary to clean.

input - The input being used to create the dictionary, possibly for sampling a new element. The list is composed of a set of column vectors.

gram - The Gram matrix of the dictionary that shows how correlated each element is with each other.