edu.stanford.nlp.maxent
Class Experiments

java.lang.Object
  extended by edu.stanford.nlp.maxent.Experiments
Direct Known Subclasses:
TaggerExperiments

public class Experiments
extends Object

This class represents the training samples. It can return statistics of them, for example the frequency of each x or y in the training data.

Author:
Kristina Toutanova

Field Summary
protected  int[] maxY
           
protected  int[] px
          px[x] holds the number of times the history x appeared in training data
protected  int[][] pxy
          pxy[x][y]=# times (x,y) occurred in training
protected  int[] py
          py[y] holds the number of times the outcome y appeared in training data
 double[][] values
          The value of classification y for x.
protected  int[][] vArray
          vArray has dimensions [numTraining][2] and holds the x and y for each training sample.
 int xSize
           
 int ySize
           
 
Constructor Summary
Experiments()
           
Experiments(int[][] vArray)
          If this constructor is used, the maximum possible class overall is found and all classes are assumed possible for all instances
Experiments(int[][] vArray, int ySize)
           
Experiments(int[][] vArray, int[] maxYs)
          The number of possible classes for each instance is contained in the array maxYs then the possible classes for x are from 0 to maxYs[x]-1
Experiments(String filename)
          The filename has format: xSizeySize x1 y1 x2 y2 ..
 
Method Summary
 void add(Experiments m)
           
 Index<IntPair> createIndex()
           
 int[] get(int index)
           
 int getNumber()
           
 int numY(int x)
           
 void print()
           
 void print(PrintFile pf)
           
 void ptilde()
           
 void ptilde(int ySize)
          When we want a pre-given number of classes.
 double ptildeX(int x)
           
 double ptildeXY(int x, int y)
           
 double ptildeY(int y)
           
 void setMaxY(int[] maxY)
           
 int size()
           
 
Methods inherited from class java.lang.Object
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait
 

Field Detail

vArray

protected int[][] vArray
vArray has dimensions [numTraining][2] and holds the x and y for each training sample. Its length is the number of data points.


px

protected int[] px
px[x] holds the number of times the history x appeared in training data


py

protected int[] py
py[y] holds the number of times the outcome y appeared in training data


maxY

protected int[] maxY

pxy

protected int[][] pxy
pxy[x][y]=# times (x,y) occurred in training


xSize

public int xSize

ySize

public int ySize

values

public double[][] values
The value of classification y for x. Used for ranking.

Constructor Detail

Experiments

public Experiments()

Experiments

public Experiments(int[][] vArray)
If this constructor is used, the maximum possible class overall is found and all classes are assumed possible for all instances


Experiments

public Experiments(int[][] vArray,
                   int[] maxYs)
The number of possible classes for each instance is contained in the array maxYs then the possible classes for x are from 0 to maxYs[x]-1


Experiments

public Experiments(int[][] vArray,
                   int ySize)

Experiments

public Experiments(String filename)
The filename has format: xSizeySize x1 y1 x2 y2 .. ..

Method Detail

createIndex

public Index<IntPair> createIndex()

add

public void add(Experiments m)

ptilde

public final void ptilde()

setMaxY

public void setMaxY(int[] maxY)

numY

public int numY(int x)

ptilde

public void ptilde(int ySize)
When we want a pre-given number of classes.


ptildeX

public double ptildeX(int x)

ptildeY

public double ptildeY(int y)

ptildeXY

public double ptildeXY(int x,
                       int y)

get

public int[] get(int index)

size

public int size()

getNumber

public int getNumber()

print

public void print()

print

public void print(PrintFile pf)


Stanford NLP Group