Package ai.libs.jaicore.ml.learningcurve.extrapolation

Class LearningCurveExtrapolator<I extends ILabeledAttributeArrayInstance<?>,D extends IOrderedLabeledAttributeArrayDataset<I,?>>

java.lang.Object

ai.libs.jaicore.ml.learningcurve.extrapolation.LearningCurveExtrapolator<I,D>

All Implemented Interfaces:

ai.libs.jaicore.basic.ILoggingCustomizable

Direct Known Subclasses:

ConfigurationLearningCurveExtrapolator

public class LearningCurveExtrapolator<I extends ILabeledAttributeArrayInstance<?>,D extends IOrderedLabeledAttributeArrayDataset<I,?>>
extends java.lang.Object
implements ai.libs.jaicore.basic.ILoggingCustomizable

Abstract class for implementing a learning curve extrapolation method with some anchor points. For each of this anchorpoints a subsample will be drawn and a classifier will be trained with this sample. Based on the points (subsample size, learner accuracy) a custom method of learning curve extrapolation can be applied.

Field Summary

Fields

Modifier and Type	Field	Description
protected D	dataset
protected LearningCurveExtrapolationMethod	extrapolationMethod
protected weka.classifiers.Classifier	learner
protected java.util.Random	random
protected ASamplingAlgorithm<I,D>	samplingAlgorithm
protected ISamplingAlgorithmFactory<I,D,? extends ASamplingAlgorithm<I,D>>	samplingAlgorithmFactory
protected D	test
protected D	train

Constructor Summary

Constructors

Constructor	Description
LearningCurveExtrapolator(LearningCurveExtrapolationMethod extrapolationMethod, weka.classifiers.Classifier learner, D dataset, double trainsplit, int[] anchorPoints, ISamplingAlgorithmFactory<I,D,? extends ASamplingAlgorithm<I,D>> samplingAlgorithmFactory, long seed)	Create a learning curve extrapolator with a subsampling factory.

Method Summary

Modifier and Type	Method	Description
LearningCurve	extrapolateLearningCurve()	Measure the learner accuracy at the given anchorpoints and extrapolate a learning curve based the results.
int[]	getAnchorPoints()
D	getDataset()
LearningCurveExtrapolationMethod	getExtrapolationMethod()
weka.classifiers.Classifier	getLearner()
java.lang.String	getLoggerName()
int[]	getTrainingTimes()
double[]	getyValues()
void	setLoggerName(java.lang.String name)

Methods inherited from class java.lang.Object

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

Field Detail

learner

protected weka.classifiers.Classifier learner

dataset

protected D extends IOrderedLabeledAttributeArrayDataset<I,?> dataset

train

protected D extends IOrderedLabeledAttributeArrayDataset<I,?> train

test

protected D extends IOrderedLabeledAttributeArrayDataset<I,?> test

samplingAlgorithmFactory

protected ISamplingAlgorithmFactory<I extends ILabeledAttributeArrayInstance<?>,D extends IOrderedLabeledAttributeArrayDataset<I,?>,? extends ASamplingAlgorithm<I extends ILabeledAttributeArrayInstance<?>,D extends IOrderedLabeledAttributeArrayDataset<I,?>>> samplingAlgorithmFactory

samplingAlgorithm

protected ASamplingAlgorithm<I extends ILabeledAttributeArrayInstance<?>,D extends IOrderedLabeledAttributeArrayDataset<I,?>> samplingAlgorithm

random

protected java.util.Random random

extrapolationMethod

protected LearningCurveExtrapolationMethod extrapolationMethod

Constructor Detail

LearningCurveExtrapolator

public LearningCurveExtrapolator(LearningCurveExtrapolationMethod extrapolationMethod,
                                 weka.classifiers.Classifier learner,
                                 D dataset,
                                 double trainsplit,
                                 int[] anchorPoints,
                                 ISamplingAlgorithmFactory<I,D,? extends ASamplingAlgorithm<I,D>> samplingAlgorithmFactory,
                                 long seed)
                          throws DatasetCreationException

Create a learning curve extrapolator with a subsampling factory.

Parameters:

extrapolationMethod - Method for extrapolating a learning curve from anchorpoints.

learner - Learning model to predict the learning curve of.

dataset - Dataset to measure evaluate the learner on.

trainsplit - Portion of the dataset, which shall be used to sample from for training.

samplingAlgorithmFactory - Subsampling algorithm factory to create a configured subsampler with.

seed - Random seed.

Throws:

DatasetCreationException

Method Detail

extrapolateLearningCurve

public LearningCurve extrapolateLearningCurve()
                                       throws InvalidAnchorPointsException,
                                              ai.libs.jaicore.basic.algorithm.exceptions.AlgorithmException,
                                              java.lang.InterruptedException

Measure the learner accuracy at the given anchorpoints and extrapolate a learning curve based the results.

Parameters:

anchorPoints - Sample sizes as anchorpoints, where the true accuracy shall be measured.

Returns:

The extrapolated learning curve.

Throws:

InvalidAnchorPointsException - The anchorpoints (amount, values, ...) are not suitable for the given learning curve extrapolation method.

ai.libs.jaicore.basic.algorithm.exceptions.AlgorithmException - An error occured during the creation of the specified anchorpoints.

java.lang.InterruptedException

getLearner

public weka.classifiers.Classifier getLearner()

getDataset

public D getDataset()

getExtrapolationMethod

public LearningCurveExtrapolationMethod getExtrapolationMethod()

getAnchorPoints

public int[] getAnchorPoints()

getyValues

public double[] getyValues()

getTrainingTimes

public int[] getTrainingTimes()

getLoggerName

public java.lang.String getLoggerName()

Specified by:

getLoggerName in interface ai.libs.jaicore.basic.ILoggingCustomizable

setLoggerName

public void setLoggerName(java.lang.String name)

Specified by:

setLoggerName in interface ai.libs.jaicore.basic.ILoggingCustomizable