autoweka

Class InstanceGenerator

java.lang.Object

autoweka.InstanceGenerator

Direct Known Subclasses:

CrossValidation, Default, RandomSubSampling

public abstract class InstanceGenerator
extends java.lang.Object

Abstract class for partitioning a dataset (consisting of training and testing data) into 'instances' that are given to the SMBO method

Nested Class Summary

Nested Classes

Modifier and Type	Class and Description
static class	InstanceGenerator.NestedArgs Struct for dealing with nested args on InstanceGenerators that are chained together.

Constructor Summary

Constructors

Constructor and Description
InstanceGenerator(InstanceGenerator generator) Clones an InstanceGenerator from another
InstanceGenerator(weka.core.Instances training, weka.core.Instances testing) Build an instance generator from the given training and testing data
InstanceGenerator(java.lang.String datasetString) Builds an InstanceGenerator from a datasetFileName, ie a zip file containing exactly two files, 'train.arff' and 'test.arff'

Method Summary

Modifier and Type	Method and Description
abstract weka.core.Instances	_getTestingFromParams(java.lang.String params)
abstract weka.core.Instances	_getTrainingFromParams(java.lang.String params)
static InstanceGenerator	create(java.lang.String className, weka.core.Instances training, weka.core.Instances testing) Creates an instance of an InstanceGenerator given a class name, training data and the testing data
static InstanceGenerator	create(java.lang.String className, java.lang.String datasetFileName) Creates an instance of an InstanceGenerator given a class name and the datasetFileName
java.util.Map<java.lang.String,java.util.Map<java.lang.String,java.lang.String>>	getAllInstanceFeatures(java.lang.String params) Some SMBO methods can leverage extra features about an Instance, this method should return a map of feature/value pairs for all instances
abstract java.util.List<java.lang.String>	getAllInstanceStrings(java.lang.String params) Gets a list of all the 'params' Strings that can be used with this InstanceGenerator.
weka.core.Instances	getTesting() Gets a copy of the testing data
weka.core.Instances	getTestingFromParams(java.lang.String params) Given a parameter string (generally in the form of a property string), get the testing data; if params is 'default', then this method returns the raw testing data
weka.core.Instances	getTraining() Gets a copy of the training data
weka.core.Instances	getTrainingFromParams(java.lang.String params) Given a parameter string (generally in the form of a property string), get the training data; if params is 'default', then this method returns the raw training data

Methods inherited from class java.lang.Object

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

Constructor Detail

InstanceGenerator

public InstanceGenerator(InstanceGenerator generator)

Clones an InstanceGenerator from another

Parameters:

generator - The instance generator.

InstanceGenerator

public InstanceGenerator(weka.core.Instances training,
                         weka.core.Instances testing)

Build an instance generator from the given training and testing data

Parameters:

training - The training instances.

testing - The testing instances.

InstanceGenerator

public InstanceGenerator(java.lang.String datasetString)

Builds an InstanceGenerator from a datasetFileName, ie a zip file containing exactly two files, 'train.arff' and 'test.arff'

Parameters:

datasetString - The string the describing the dataset.

Method Detail

getTrainingFromParams

public final weka.core.Instances getTrainingFromParams(java.lang.String params)

Given a parameter string (generally in the form of a property string), get the training data; if params is 'default', then this method returns the raw training data

Parameters:

params - The parameter string.

Returns:

The training instances.

getTestingFromParams

public final weka.core.Instances getTestingFromParams(java.lang.String params)

Given a parameter string (generally in the form of a property string), get the testing data; if params is 'default', then this method returns the raw testing data

Parameters:

params - The parameter string.

Returns:

The testing instances.

_getTrainingFromParams

public abstract weka.core.Instances _getTrainingFromParams(java.lang.String params)

_getTestingFromParams

public abstract weka.core.Instances _getTestingFromParams(java.lang.String params)

getTraining

public weka.core.Instances getTraining()

Gets a copy of the training data

Returns:

The training instances.

getTesting

public weka.core.Instances getTesting()

Gets a copy of the testing data

Returns:

The testing instances.

getAllInstanceStrings

public abstract java.util.List<java.lang.String> getAllInstanceStrings(java.lang.String params)

Gets a list of all the 'params' Strings that can be used with this InstanceGenerator. For example, the for 10-fold Cross Validation this method would return 10 Strings defining each fold

Parameters:

params - The parameters.

Returns:

The list of parameters.

getAllInstanceFeatures

public java.util.Map<java.lang.String,java.util.Map<java.lang.String,java.lang.String>> getAllInstanceFeatures(java.lang.String params)

Some SMBO methods can leverage extra features about an Instance, this method should return a map of feature/value pairs for all instances

Parameters:

params - The parameters.

Returns:

The map of feature/value pairs.

create

public static InstanceGenerator create(java.lang.String className,
                                       java.lang.String datasetFileName)

Creates an instance of an InstanceGenerator given a class name and the datasetFileName

Parameters:

className - The name of the class.

datasetFileName - The name of the dataset file.

Returns:

The instance generator.

create

public static InstanceGenerator create(java.lang.String className,
                                       weka.core.Instances training,
                                       weka.core.Instances testing)

Creates an instance of an InstanceGenerator given a class name, training data and the testing data

Parameters:

className - The name of the class.

training - The training instances.

testing - The testing instances.

Returns:

The instance generator.