Package ciir.umass.edu.features

Class FeatureManager

java.lang.Object

ciir.umass.edu.features.FeatureManager

public class FeatureManager
extends java.lang.Object

Constructor Summary

Constructors

Constructor	Description
FeatureManager()

Method Summary

Modifier and Type	Method	Description
static int[]	getFeatureFromSampleVector(java.util.List<RankList> samples)	Obtain all features present in a sample set.
static void	main(java.lang.String[] args)
static void	prepareCV(java.util.List<RankList> samples, int nFold, float tvs, java.util.List<java.util.List<RankList>> trainingData, java.util.List<java.util.List<RankList>> validationData, java.util.List<java.util.List<RankList>> testData)	Split the input sample set into k chunks (folds) of roughly equal size and create train/test data for each fold.
static void	prepareCV(java.util.List<RankList> samples, int nFold, java.util.List<java.util.List<RankList>> trainingData, java.util.List<java.util.List<RankList>> testData)	Split the input sample set into k chunks (folds) of roughly equal size and create train/test data for each fold.
static void	prepareSplit(java.util.List<RankList> samples, double percentTrain, java.util.List<RankList> trainingData, java.util.List<RankList> testData)	Split the input sample set into 2 chunks: one for training and one for either validation or testing
static void	printQueriesForSplit(java.lang.String name, java.util.List<java.util.List<RankList>> split)
static int[]	readFeature(java.lang.String featureDefFile)	Read features specified in an input feature file.
static java.util.List<RankList>	readInput(java.lang.String inputFile)	Read a set of rankings from a single file.
static java.util.List<RankList>	readInput(java.lang.String inputFile, boolean mustHaveRelDoc, boolean useSparseRepresentation)	Read a set of rankings from a single file.
static java.util.List<RankList>	readInput(java.util.List<java.lang.String> inputFiles)	Read sets of rankings from multiple files.
static void	save(java.util.List<RankList> samples, java.lang.String outputFile)	Save a sample set to file

Methods inherited from class java.lang.Object

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

Constructor Detail

FeatureManager

public FeatureManager()

Method Detail

main

public static void main(java.lang.String[] args)

Parameters:

args -

readInput

public static java.util.List<RankList> readInput(java.lang.String inputFile)

Read a set of rankings from a single file.

Parameters:

inputFile -

Returns:

readInput

public static java.util.List<RankList> readInput(java.lang.String inputFile,
                                                 boolean mustHaveRelDoc,
                                                 boolean useSparseRepresentation)

Read a set of rankings from a single file.

Parameters:

inputFile -

mustHaveRelDoc -

useSparseRepresentation -

Returns:

readInput

public static java.util.List<RankList> readInput(java.util.List<java.lang.String> inputFiles)

Read sets of rankings from multiple files. Then merge them altogether into a single ranking.

Parameters:

inputFiles -

Returns:

readFeature

public static int[] readFeature(java.lang.String featureDefFile)

Read features specified in an input feature file. Expecting one feature per line.

Parameters:

featureDefFile -

Returns:

getFeatureFromSampleVector

public static int[] getFeatureFromSampleVector(java.util.List<RankList> samples)

Obtain all features present in a sample set.

Parameters:

samples -

Returns:

prepareCV

public static void prepareCV(java.util.List<RankList> samples,
                             int nFold,
                             java.util.List<java.util.List<RankList>> trainingData,
                             java.util.List<java.util.List<RankList>> testData)

Split the input sample set into k chunks (folds) of roughly equal size and create train/test data for each fold. Note that NO randomization is done. If you want to randomly split the data, make sure that you randomize the order in the input samples prior to calling this function.

Parameters:

samples -

nFold -

trainingData -

testData -

prepareCV

public static void prepareCV(java.util.List<RankList> samples,
                             int nFold,
                             float tvs,
                             java.util.List<java.util.List<RankList>> trainingData,
                             java.util.List<java.util.List<RankList>> validationData,
                             java.util.List<java.util.List<RankList>> testData)

Split the input sample set into k chunks (folds) of roughly equal size and create train/test data for each fold. Then it further splits the training data in each fold into train and validation. Note that NO randomization is done. If you want to randomly split the data, make sure that you randomize the order in the input samples prior to calling this function.

Parameters:

samples -

nFold -

tvs - Train/validation split ratio

trainingData -

validationData -

testData -

printQueriesForSplit

public static void printQueriesForSplit(java.lang.String name,
                                        java.util.List<java.util.List<RankList>> split)

prepareSplit

public static void prepareSplit(java.util.List<RankList> samples,
                                double percentTrain,
                                java.util.List<RankList> trainingData,
                                java.util.List<RankList> testData)

Split the input sample set into 2 chunks: one for training and one for either validation or testing

Parameters:

samples -

percentTrain - The percentage of data used for training

trainingData -

testData -

save

public static void save(java.util.List<RankList> samples,
                        java.lang.String outputFile)

Save a sample set to file

Parameters:

samples -

outputFile -