JMSLTM Numerical Library 5.0.1

com.imsl.datamining.neural
Class UnsupervisedNominalFilter

java.lang.Object
  extended by com.imsl.datamining.neural.UnsupervisedNominalFilter
All Implemented Interfaces:
Serializable

public class UnsupervisedNominalFilter
extends Object
implements Serializable

Converts nominal data into a series of binary encoded columns for input to a neural network. It also reverses the aforementioned encoding, accepting binary encoded data and returns an array of integers representing the classes for a nominal variable.

Binary Encoding

Method encode can be used to apply binary encoding. Referring to the result as z, binary encoding takes each category in the nominal variable x[], and creates a column in z containing all zeros and ones. A value of zero indicates that this category was not present and a value of one indicates that it is present.

For example, if x[]={2, 1, 3, 4, 2, 4} then nClasses=4, and

z = begin{array} {rrrr}
 0 & 1 & 0 & 0 \
 1 & 0 & 0 & 0 \
 0 & 0 & 1 & 0 \
 0 & 0 & 0 & 1 \
 0 & 1 & 0 & 0 \
 0 & 0 & 0 & 1
 end{array}

Notice that the number of columns in the result, z, is equal to the number of distinct classes in x. The number of rows in z is equal to the length of x.

Binary Decoding

Unfiltering can be performed using the method decode. In this case, z is the input, and we refer to x as the output. Binary unfiltering takes binary representation in z, and returns the appropriate class in x.

For example, if a row in z equals {0, 1, 0, 0}, then the return value from decode would be 2 for that row. If a row in z equals {1, 0, 0, 0}, then the return value from decode would be 1 for that row. Notice these are the same values as the first two elements of the original x[] because classes are numbered sequentially from 1 to nClasses. This ensures that the results of decode are associated with the i-th class in x[].

See Also:
Example, Serialized Form

Constructor Summary
UnsupervisedNominalFilter(int nClasses)
          Constructor for UnsupervisedNominalFilter.
 
Method Summary
 int decode(int[] z)
          Decodes a binary encoded array into its nominal category.
 int[] decode(int[][] z)
          Decodes a matrix representing the binary encoded columns of the nominal variable.
 int[] encode(int x)
          Apply forward encoding to a value.
 int[][] encode(int[] x)
          Encodes class data prior to its use in neural network training.
 int getNumberOfClasses()
          Retrieves the number of classes in the nominal variable.
 
Methods inherited from class java.lang.Object
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait
 

Constructor Detail

UnsupervisedNominalFilter

public UnsupervisedNominalFilter(int nClasses)
Constructor for UnsupervisedNominalFilter.

Parameters:
nClasses - An int specifying the number of categories in the nominal variable to be filtered.
Method Detail

decode

public int decode(int[] z)
Decodes a binary encoded array into its nominal category. This is the inverse of the encode(int) method.

Parameters:
z - An int array containing the data to be decoded.
Returns:
An int containing the number associated with the category encoded in z.

decode

public int[] decode(int[][] z)
Decodes a matrix representing the binary encoded columns of the nominal variable. This is the inverse of the encode(int[]) method.

Parameters:
z - An int matrix containing the data to be decoded.
Returns:
An int array containing the decoded data.

encode

public int[] encode(int x)
Apply forward encoding to a value.

Parameters:
x - An int containing the value to be encoding. Class number must be in the range 1 to nClasses.
Returns:
An int array containing the encoded data.

encode

public int[][] encode(int[] x)
Encodes class data prior to its use in neural network training.

Parameters:
x - An int array containing the data to be encoded. Class number must be in the range 1 to nClasses.
Returns:
An int matrix containing the encoded data.

getNumberOfClasses

public int getNumberOfClasses()
Retrieves the number of classes in the nominal variable.

Returns:
An int containing the number of classes in the nominal variable.

JMSLTM Numerical Library 5.0.1

Copyright © 1970-2008 Visual Numerics, Inc.
Built July 8 2008.