ai.libs.mlplan.core

Class AbstractMLPlanBuilder

java.lang.Object

ai.libs.mlplan.core.AbstractMLPlanBuilder

All Implemented Interfaces:

ai.libs.jaicore.basic.ILoggingCustomizable, IMLPlanBuilder

Direct Known Subclasses:

AbstractMLPlanSingleLabelBuilder, MLPlanMekaBuilder

public abstract class AbstractMLPlanBuilder
extends java.lang.Object
implements IMLPlanBuilder, ai.libs.jaicore.basic.ILoggingCustomizable

The MLPlanBuilder helps to easily configure and initialize ML-Plan with specific parameter settings. For convenient use, the MLPlanBuilder also offers methods for initializing ML-Plan with default configuration to use ML-Plan for single label classification in combination with WEKA or scikit-learn or for multi-label classification in combination with MEKA and consequently with WEKA (for baselearners of multi-label reduction strategies).

Constructor Summary

Constructors

Modifier	Constructor and Description
protected	AbstractMLPlanBuilder()

Method Summary

Modifier and Type	Method and Description
MLPlan	build() Builds an ML-Plan object with the dataset provided earlier to this builder.
MLPlan	build(weka.core.Instances dataset) Builds an ML-Plan object for the given dataset as input.
static MLPlanMekaBuilder	forMeka()
static MLPlanSKLearnBuilder	forSKLearn()
static MLPlanWekaBuilder	forWeka()
MLPlanClassifierConfig	getAlgorithmConfig()
ai.libs.jaicore.basic.TimeOut	getCandidateEvaluationTimeOut()
PipelineEvaluator	getClassifierEvaluationInSearchPhase(weka.core.Instances data, int seed, int fullDatasetSize)
PipelineEvaluator	getClassifierEvaluationInSelectionPhase(weka.core.Instances data, int seed)
IClassifierFactory	getClassifierFactory()
java.util.Collection<ai.libs.hasco.model.Component>	getComponents()
PerformanceDBAdapter	getDBAdapter()
ai.libs.hasco.core.HASCOFactory	getHASCOFactory()
java.lang.String	getLoggerName()
ai.libs.jaicore.basic.TimeOut	getNodeEvaluationTimeOut()
java.lang.String	getPerformanceMeasureName()
java.lang.String	getRequestedInterface()
protected ai.libs.jaicore.ml.evaluation.evaluators.weka.factory.IClassifierEvaluatorFactory	getSearchEvaluatorFactory()
ai.libs.jaicore.ml.weka.dataset.splitter.IDatasetSplitter	getSearchSelectionDatasetSplitter()
java.io.File	getSearchSpaceConfigFile()
protected ai.libs.jaicore.ml.evaluation.evaluators.weka.factory.IClassifierEvaluatorFactory	getSelectionEvaluatorFactory()
ai.libs.jaicore.basic.TimeOut	getTimeOut()
boolean	getUseCache()
void	prepareNodeEvaluatorInFactoryWithData(weka.core.Instances data)
void	setLoggerName(java.lang.String name)
void	setPerformanceMeasureName(java.lang.String name) Sets the name of the performance measure that is used.
AbstractMLPlanBuilder	withAlgorithmConfig(MLPlanClassifierConfig config) Loads the MLPlanClassifierConfig with default values and replaces all properties according to the properties defined in the given config file.
AbstractMLPlanBuilder	withAlgorithmConfigFile(java.io.File algorithmConfigFile) Loads the MLPlanClassifierConfig with default values and replaces all properties according to the properties defined in the given config file.
AbstractMLPlanBuilder	withCandidateEvaluationTimeOut(ai.libs.jaicore.basic.TimeOut timeout)
AbstractMLPlanBuilder	withClassifierFactory(IClassifierFactory classifierFactory) Set the classifier factory that translates CompositionInstance objects to classifiers that can be evaluated.
AbstractMLPlanBuilder	withDataset(weka.core.Instances dataset) Set the data for which ML-Plan is supposed to find the best pipeline.
AbstractMLPlanBuilder	withDatasetSplitterForSearchSelectionSplit(ai.libs.jaicore.ml.weka.dataset.splitter.IDatasetSplitter datasetSplitter) Set the dataset splitter that is used for generating the holdout data portion that is put aside during search.
AbstractMLPlanBuilder	withNodeEvaluationTimeOut(ai.libs.jaicore.basic.TimeOut timeout)
AbstractMLPlanBuilder	withNumCpus(int numCpus) Sets the number of cpus that may be used by ML-Plan.
AbstractMLPlanBuilder	withPreferredComponentsFile(java.io.File preferredComponentsFile) Creates a preferred node evaluator that can be used to prefer components over other components.
AbstractMLPlanBuilder	withPreferredNodeEvaluator(jaicore.search.algorithms.standard.bestfirst.nodeevaluation.INodeEvaluator<ai.libs.jaicore.planning.hierarchical.algorithms.forwarddecomposition.graphgenerators.tfd.TFDNode,java.lang.Double> preferredNodeEvaluator) This ADDs a new preferred node evaluator; requires that the search will be a best-first search.
AbstractMLPlanBuilder	withRandomCompletionBasedBestFirstSearch()
AbstractMLPlanBuilder	withRequestedInterface(java.lang.String requestedInterface)
AbstractMLPlanBuilder	withSearchFactory(jaicore.search.core.interfaces.IOptimalPathInORGraphSearchFactory searchFactory, ai.libs.jaicore.basic.algorithm.reduction.AlgorithmicProblemReduction transformer)
void	withSearchPhaseEvaluatorFactory(ai.libs.jaicore.ml.evaluation.evaluators.weka.factory.IClassifierEvaluatorFactory evaluatorFactory) Sets the evaluator factory for the search phase.
AbstractMLPlanBuilder	withSearchSpaceConfigFile(java.io.File searchSpaceConfig) Specify the search space in which ML-Plan is required to work.
AbstractMLPlanBuilder	withSelectionPhaseEvaluatorFactory(ai.libs.jaicore.ml.evaluation.evaluators.weka.factory.MonteCarloCrossValidationEvaluatorFactory evaluatorFactory) Sets the evaluator factory for the selection phase.
AbstractMLPlanBuilder	withTimeOut(ai.libs.jaicore.basic.TimeOut timeout)

Methods inherited from class java.lang.Object

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

Constructor Detail

AbstractMLPlanBuilder

protected AbstractMLPlanBuilder()

Method Detail

forSKLearn

public static MLPlanSKLearnBuilder forSKLearn()
                                       throws java.io.IOException

Throws:

java.io.IOException

forWeka

public static MLPlanWekaBuilder forWeka()
                                 throws java.io.IOException

Throws:

java.io.IOException

forMeka

public static MLPlanMekaBuilder forMeka()
                                 throws java.io.IOException

Throws:

java.io.IOException

withPreferredNodeEvaluator

public AbstractMLPlanBuilder withPreferredNodeEvaluator(jaicore.search.algorithms.standard.bestfirst.nodeevaluation.INodeEvaluator<ai.libs.jaicore.planning.hierarchical.algorithms.forwarddecomposition.graphgenerators.tfd.TFDNode,java.lang.Double> preferredNodeEvaluator)

This ADDs a new preferred node evaluator; requires that the search will be a best-first search. It is possible to specify several preferred node evaluators, which will be ordered by the order in which they are specified. The latest given evaluator is the most preferred one.

Parameters:

preferredNodeEvaluator -

Returns:

withSearchFactory

public AbstractMLPlanBuilder withSearchFactory(jaicore.search.core.interfaces.IOptimalPathInORGraphSearchFactory searchFactory,
                                               ai.libs.jaicore.basic.algorithm.reduction.AlgorithmicProblemReduction transformer)

withRandomCompletionBasedBestFirstSearch

public AbstractMLPlanBuilder withRandomCompletionBasedBestFirstSearch()

getComponents

public java.util.Collection<ai.libs.hasco.model.Component> getComponents()
                                                                  throws java.io.IOException

Throws:

java.io.IOException

withAlgorithmConfigFile

public AbstractMLPlanBuilder withAlgorithmConfigFile(java.io.File algorithmConfigFile)

Loads the MLPlanClassifierConfig with default values and replaces all properties according to the properties defined in the given config file.

Parameters:

algorithmConfigFile - The file specifying the property values to replace the default configuration.

Returns:

The MLPlanBuilder object.

Throws:

java.io.IOException - An IOException is thrown if there are issues reading the config file.

withAlgorithmConfig

public AbstractMLPlanBuilder withAlgorithmConfig(MLPlanClassifierConfig config)

Loads the MLPlanClassifierConfig with default values and replaces all properties according to the properties defined in the given config file.

Parameters:

config - The algorithm configuration.

Returns:

The MLPlanBuilder object.

Throws:

java.io.IOException - An IOException is thrown if there are issues reading the config file.

withPreferredComponentsFile

public AbstractMLPlanBuilder withPreferredComponentsFile(java.io.File preferredComponentsFile)
                                                  throws java.io.IOException

Creates a preferred node evaluator that can be used to prefer components over other components.

Parameters:

preferredComponentsFile - The file containing a priority list of component names.

Returns:

The builder object.

Throws:

java.io.IOException - Thrown if a problem occurs while trying to read the file containing the priority list.

setPerformanceMeasureName

public void setPerformanceMeasureName(java.lang.String name)

Sets the name of the performance measure that is used.

Parameters:

name - The name of the performance measure.

withDataset

public AbstractMLPlanBuilder withDataset(weka.core.Instances dataset)

Set the data for which ML-Plan is supposed to find the best pipeline.

Parameters:

dataset - The dataset for which ML-Plan is to be run.

Returns:

The builder object.

withSearchSpaceConfigFile

public AbstractMLPlanBuilder withSearchSpaceConfigFile(java.io.File searchSpaceConfig)
                                                throws java.io.IOException

Specify the search space in which ML-Plan is required to work.

Parameters:

searchSpaceConfig - The file of the search space configuration.

Returns:

The builder object.

Throws:

java.io.IOException - Thrown if the given file does not exist.

withClassifierFactory

public AbstractMLPlanBuilder withClassifierFactory(IClassifierFactory classifierFactory)

Set the classifier factory that translates CompositionInstance objects to classifiers that can be evaluated.

Parameters:

classifierFactory - The classifier factory to be used to translate CompositionInstance objects to classifiers.

Returns:

The builder object.

withDatasetSplitterForSearchSelectionSplit

public AbstractMLPlanBuilder withDatasetSplitterForSearchSelectionSplit(ai.libs.jaicore.ml.weka.dataset.splitter.IDatasetSplitter datasetSplitter)

Set the dataset splitter that is used for generating the holdout data portion that is put aside during search.

Parameters:

datasetSplitter - The dataset splitter to be used.

Returns:

The builder obect.

withRequestedInterface

public AbstractMLPlanBuilder withRequestedInterface(java.lang.String requestedInterface)

withTimeOut

public AbstractMLPlanBuilder withTimeOut(ai.libs.jaicore.basic.TimeOut timeout)

Parameters:

timeout - The timeout for ML-Plan to search for the best classifier.

Returns:

The builder object.

getTimeOut

public ai.libs.jaicore.basic.TimeOut getTimeOut()

Returns:

The timeout for ML-Plan to search for the best classifier.

withNodeEvaluationTimeOut

public AbstractMLPlanBuilder withNodeEvaluationTimeOut(ai.libs.jaicore.basic.TimeOut timeout)

Parameters:

timeout - The timeout for a single candidate evaluation.

Returns:

The builder object.

getNodeEvaluationTimeOut

public ai.libs.jaicore.basic.TimeOut getNodeEvaluationTimeOut()

Returns:

The timeout for ML-Plan to search for the best classifier.

withCandidateEvaluationTimeOut

public AbstractMLPlanBuilder withCandidateEvaluationTimeOut(ai.libs.jaicore.basic.TimeOut timeout)

Parameters:

timeout - The timeout for a single candidate evaluation.

Returns:

The builder object.

getCandidateEvaluationTimeOut

public ai.libs.jaicore.basic.TimeOut getCandidateEvaluationTimeOut()

Returns:

The timeout for ML-Plan to search for the best classifier.

getClassifierEvaluationInSearchPhase

public PipelineEvaluator getClassifierEvaluationInSearchPhase(weka.core.Instances data,
                                                              int seed,
                                                              int fullDatasetSize)
                                                       throws ai.libs.jaicore.ml.evaluation.evaluators.weka.factory.ClassifierEvaluatorConstructionFailedException

Specified by:

getClassifierEvaluationInSearchPhase in interface IMLPlanBuilder

Throws:

ai.libs.jaicore.ml.evaluation.evaluators.weka.factory.ClassifierEvaluatorConstructionFailedException

getClassifierEvaluationInSelectionPhase

public PipelineEvaluator getClassifierEvaluationInSelectionPhase(weka.core.Instances data,
                                                                 int seed)
                                                          throws ai.libs.jaicore.ml.evaluation.evaluators.weka.factory.ClassifierEvaluatorConstructionFailedException

Specified by:

getClassifierEvaluationInSelectionPhase in interface IMLPlanBuilder

Throws:

ai.libs.jaicore.ml.evaluation.evaluators.weka.factory.ClassifierEvaluatorConstructionFailedException

withSearchPhaseEvaluatorFactory

public void withSearchPhaseEvaluatorFactory(ai.libs.jaicore.ml.evaluation.evaluators.weka.factory.IClassifierEvaluatorFactory evaluatorFactory)

Sets the evaluator factory for the search phase.

Parameters:

evaluatorFactory - The evaluator factory for the search phase.

getSearchEvaluatorFactory

protected ai.libs.jaicore.ml.evaluation.evaluators.weka.factory.IClassifierEvaluatorFactory getSearchEvaluatorFactory()

Returns:

The factory for the classifier evaluator of the search phase.

withSelectionPhaseEvaluatorFactory

public AbstractMLPlanBuilder withSelectionPhaseEvaluatorFactory(ai.libs.jaicore.ml.evaluation.evaluators.weka.factory.MonteCarloCrossValidationEvaluatorFactory evaluatorFactory)

Sets the evaluator factory for the selection phase.

Parameters:

evaluatorFactory - The evaluator factory for the selection phase.

Returns:

The builder object.

withNumCpus

public AbstractMLPlanBuilder withNumCpus(int numCpus)

Sets the number of cpus that may be used by ML-Plan.

Parameters:

numCpus - The number of cpus to use.

Returns:

The builder object.

getSelectionEvaluatorFactory

protected ai.libs.jaicore.ml.evaluation.evaluators.weka.factory.IClassifierEvaluatorFactory getSelectionEvaluatorFactory()

Returns:

The factory for the classifier evaluator of the selection phase.

getPerformanceMeasureName

public java.lang.String getPerformanceMeasureName()

Specified by:

getPerformanceMeasureName in interface IMLPlanBuilder

getHASCOFactory

public ai.libs.hasco.core.HASCOFactory getHASCOFactory()

Specified by:

getHASCOFactory in interface IMLPlanBuilder

getClassifierFactory

public IClassifierFactory getClassifierFactory()

Specified by:

getClassifierFactory in interface IMLPlanBuilder

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

getRequestedInterface

public java.lang.String getRequestedInterface()

Specified by:

getRequestedInterface in interface IMLPlanBuilder

getSearchSelectionDatasetSplitter

public ai.libs.jaicore.ml.weka.dataset.splitter.IDatasetSplitter getSearchSelectionDatasetSplitter()

Specified by:

getSearchSelectionDatasetSplitter in interface IMLPlanBuilder

getSearchSpaceConfigFile

public java.io.File getSearchSpaceConfigFile()

Specified by:

getSearchSpaceConfigFile in interface IMLPlanBuilder

getAlgorithmConfig

public MLPlanClassifierConfig getAlgorithmConfig()

Specified by:

getAlgorithmConfig in interface IMLPlanBuilder

getUseCache

public boolean getUseCache()

Specified by:

getUseCache in interface IMLPlanBuilder

getDBAdapter

public PerformanceDBAdapter getDBAdapter()

Specified by:

getDBAdapter in interface IMLPlanBuilder

prepareNodeEvaluatorInFactoryWithData

public void prepareNodeEvaluatorInFactoryWithData(weka.core.Instances data)

Specified by:

prepareNodeEvaluatorInFactoryWithData in interface IMLPlanBuilder

build

public MLPlan build(weka.core.Instances dataset)

Builds an ML-Plan object for the given dataset as input.

Parameters:

dataset - The dataset for which an ML-Plan object is to be built.

Returns:

The ML-Plan object configured with this builder.

build

public MLPlan build()

Builds an ML-Plan object with the dataset provided earlier to this builder.

Returns:

The ML-Plan object configured with this builder.