edu.illinois.cs.cogcomp.core.experiments

Class CrossValidationHelper<T>

java.lang.Object

edu.illinois.cs.cogcomp.core.experiments.CrossValidationHelper<T>

Type Parameters:

T - The type of the training examples

public class CrossValidationHelper<T>
extends Object

Author:

Vivek Srikumar

Jan 30, 2009

Field Summary

Fields

Modifier and Type	Field and Description
protected int	numFolds

Constructor Summary

Constructors

Constructor and Description
CrossValidationHelper(IExperimentFactory<T> experiment) By default, do five fold cross validation
CrossValidationHelper(int numFolds, IExperimentFactory<T> experiment)
CrossValidationHelper(long timeoutSeconds, int numThreads, int numFolds, IExperimentFactory<T> experimentFactory)

Method Summary

Modifier and Type	Method and Description
double	doCrossValidation(Iterable<T> data, int dataSize) NOTE: This does not take care of shuffling or randomizing the data.
OneVariableStats	getStats()
protected List<List<T>>	splitData(Iterable<T> data, int dataSize) Splits the data into numFolds parts.

Methods inherited from class java.lang.Object

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

Field Detail

numFolds

protected final int numFolds

Constructor Detail

CrossValidationHelper

public CrossValidationHelper(IExperimentFactory<T> experiment)

By default, do five fold cross validation

Parameters:

experiment - One run of the experiment ( test + train)

CrossValidationHelper

public CrossValidationHelper(int numFolds,
                             IExperimentFactory<T> experiment)

Parameters:

numFolds - number of folds

experiment - one run of the experiment (test + train)

CrossValidationHelper

public CrossValidationHelper(long timeoutSeconds,
                             int numThreads,
                             int numFolds,
                             IExperimentFactory<T> experimentFactory)

Method Detail

doCrossValidation

public double doCrossValidation(Iterable<T> data,
                                int dataSize)
                         throws InterruptedException,
                                ExecutionException

NOTE: This does not take care of shuffling or randomizing the data.

Throws:

InterruptedException

ExecutionException

getStats

public OneVariableStats getStats()

splitData

protected List<List<T>> splitData(Iterable<T> data,
                                  int dataSize)

Splits the data into numFolds parts.

Note: This splits the data into K folds uniformly. If the classes are not equally distributed, then this is wrong. Instead, override this to do a stratified split, so that the split proportions are maintained.