edu.stanford.nlp.maxent.iis
Class BinaryFeature

java.lang.Object
  |
  +--edu.stanford.nlp.maxent.iis.Feature
        |
        +--edu.stanford.nlp.maxent.iis.BinaryFeature

public class BinaryFeature

extends Feature

This is used when only binary features are needed. We don't need to keep the array of values valuesI any more (unlike the parent class Feature)

Field Summary

Fields inherited from class edu.stanford.nlp.maxent.iis.Feature

hashValues, indexedValues, sum, valuesI

Constructor Summary

BinaryFeature()

BinaryFeature(Experiments e, double[] vals)

BinaryFeature(Experiments domain, double[][] vals)

BinaryFeature(Experiments e, int[] indVals)

Method Summary

double	ftiled()
void	getIndexed() Convert the setValues into an array of indexes of non-zero pairs indexedValues.
double	getVal(int index) Get the value at the index-ed non zero value pair (x,y)
double	getVal(int x, int y) This does sequential search in indexedValues.
void	getValues()
boolean	isEmpty()
void	print() Print put the pairs for which the feature is true

Methods inherited from class edu.stanford.nlp.maxent.iis.Feature

getX, getY, initHashVals, len, print, read, save, setSum, sumValues

Methods inherited from class java.lang.Object

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

Constructor Detail

BinaryFeature

public BinaryFeature()

BinaryFeature

public BinaryFeature(Experiments e,
                     double[] vals)

Parameters:

vals - will have length equal to the number of training samples (x,y) vals contains only 0s and 1s

BinaryFeature

public BinaryFeature(Experiments e,
                     int[] indVals)

BinaryFeature

public BinaryFeature(Experiments domain,
                     double[][] vals)

Method Detail

isEmpty

public boolean isEmpty()

Overrides:

isEmpty in class Feature

Returns:

true if the feature does not have a non-zero value at any data point

getValues

public void getValues()

getIndexed

public void getIndexed()

Convert the setValues into an array of indexes of non-zero pairs indexedValues.

print

public void print()

Print put the pairs for which the feature is true

Overrides:

print in class Feature

getVal

public double getVal(int index)

Description copied from class: Feature

Get the value at the index-ed non zero value pair (x,y)

Overrides:

getVal in class Feature

getVal

public double getVal(int x,
                     int y)

This does sequential search in indexedValues. It is slow.

Overrides:

getVal in class Feature

Parameters:

x - the history Id

y - the outcome Id

Returns:

the value of the feature at the point (x,y)

ftiled

public double ftiled()

Overrides:

ftiled in class Feature

Returns:

the empirical expectation of the feature