org.tribuo.regression.liblinear

Class LibLinearRegressionTrainer

java.lang.Object

org.tribuo.common.liblinear.LibLinearTrainer<Regressor>

org.tribuo.regression.liblinear.LibLinearRegressionTrainer

All Implemented Interfaces:

com.oracle.labs.mlrg.olcut.config.Configurable, com.oracle.labs.mlrg.olcut.provenance.Provenancable<TrainerProvenance>, Trainer<Regressor>

public class LibLinearRegressionTrainer
extends LibLinearTrainer<Regressor>

A Trainer which wraps a liblinear-java regression trainer.

This generates an independent liblinear model for each regression dimension.

Note the train method is synchronized on LibLinearTrainer.class due to a global RNG in liblinear-java. This is insufficient to ensure reproducibility if liblinear-java is used directly in the same JVM as Tribuo, but avoids locking on classes Tribuo does not control.

See:

 Fan RE, Chang KW, Hsieh CJ, Wang XR, Lin CJ.
 "LIBLINEAR: A library for Large Linear Classification"
 Journal of Machine Learning Research, 2008.
 

and for the original algorithm:

 Cortes C, Vapnik V.
 "Support-Vector Networks"
 Machine Learning, 1995.
 

Field Summary

Fields inherited from class org.tribuo.common.liblinear.LibLinearTrainer

cost, epsilon, libLinearParams, maxIterations, terminationCriterion, trainerType

Fields inherited from interface org.tribuo.Trainer

DEFAULT_SEED

Constructor Summary

Constructors

Constructor and Description
LibLinearRegressionTrainer() Creates a trainer using the default values (L2R_L2LOSS_SVR, 1, 1000, 0.1, 0.1).
LibLinearRegressionTrainer(LinearRegressionType trainerType) Creates a trainer for a LibLinear regression model.
LibLinearRegressionTrainer(LinearRegressionType trainerType, double cost, int maxIterations, double terminationCriterion, double epsilon) Creates a trainer for a LibLinear regression model.

Method Summary

Modifier and Type	Method and Description
protected LibLinearModel<Regressor>	createModel(ModelProvenance provenance, ImmutableFeatureMap featureIDMap, ImmutableOutputInfo<Regressor> outputIDInfo, List<de.bwaldvogel.liblinear.Model> models) Construct the appropriate subtype of LibLinearModel for the prediction task.
protected com.oracle.labs.mlrg.olcut.util.Pair<de.bwaldvogel.liblinear.FeatureNode[][],double[][]>	extractData(Dataset<Regressor> data, ImmutableOutputInfo<Regressor> outputInfo, ImmutableFeatureMap featureMap) Extracts the features and Outputs in LibLinear's format.
void	postConfig() Used by the OLCUT configuration system, and should not be called by external code.
protected List<de.bwaldvogel.liblinear.Model>	trainModels(de.bwaldvogel.liblinear.Parameter curParams, int numFeatures, de.bwaldvogel.liblinear.FeatureNode[][] features, double[][] outputs) Train all the liblinear instances necessary for this dataset.

Methods inherited from class org.tribuo.common.liblinear.LibLinearTrainer

exampleToNodes, getInvocationCount, getProvenance, setupParameters, toString, train, train

Methods inherited from class java.lang.Object

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

Constructor Detail

LibLinearRegressionTrainer

public LibLinearRegressionTrainer()

Creates a trainer using the default values (L2R_L2LOSS_SVR, 1, 1000, 0.1, 0.1).

LibLinearRegressionTrainer

public LibLinearRegressionTrainer(LinearRegressionType trainerType)

Creates a trainer for a LibLinear regression model.

Uses the default values of cost = 1.0, maxIterations = 1000, terminationCriterion = 0.1, epsilon = 0.1.

Parameters:

trainerType - Loss function and optimisation method.

LibLinearRegressionTrainer

public LibLinearRegressionTrainer(LinearRegressionType trainerType,
                                  double cost,
                                  int maxIterations,
                                  double terminationCriterion,
                                  double epsilon)

Creates a trainer for a LibLinear regression model.

Parameters:

trainerType - Loss function and optimisation method combination.

cost - Cost penalty for each incorrectly classified training point.

maxIterations - The maximum number of dataset iterations.

terminationCriterion - How close does the optimisation function need to be before terminating that subproblem (usually set to 0.1).

epsilon - The insensitivity of the regression loss to small differences.

Method Detail

postConfig

public void postConfig()

Used by the OLCUT configuration system, and should not be called by external code.

Specified by:

postConfig in interface com.oracle.labs.mlrg.olcut.config.Configurable

Overrides:

postConfig in class LibLinearTrainer<Regressor>

trainModels

protected List<de.bwaldvogel.liblinear.Model> trainModels(de.bwaldvogel.liblinear.Parameter curParams,
                                                          int numFeatures,
                                                          de.bwaldvogel.liblinear.FeatureNode[][] features,
                                                          double[][] outputs)

Description copied from class: LibLinearTrainer

Train all the liblinear instances necessary for this dataset.

Specified by:

trainModels in class LibLinearTrainer<Regressor>

Parameters:

curParams - The LibLinear parameters.

numFeatures - The number of features in this dataset.

features - The features themselves.

outputs - The outputs.

Returns:

A list of liblinear models.

createModel

protected LibLinearModel<Regressor> createModel(ModelProvenance provenance,
                                                ImmutableFeatureMap featureIDMap,
                                                ImmutableOutputInfo<Regressor> outputIDInfo,
                                                List<de.bwaldvogel.liblinear.Model> models)

Description copied from class: LibLinearTrainer

Construct the appropriate subtype of LibLinearModel for the prediction task.

Specified by:

createModel in class LibLinearTrainer<Regressor>

Parameters:

provenance - The model provenance.

featureIDMap - The feature id map.

outputIDInfo - The output id info.

models - The list of linear models.

Returns:

An implementation of LibLinearModel.

extractData

protected com.oracle.labs.mlrg.olcut.util.Pair<de.bwaldvogel.liblinear.FeatureNode[][],double[][]> extractData(Dataset<Regressor> data,
                                                                                                               ImmutableOutputInfo<Regressor> outputInfo,
                                                                                                               ImmutableFeatureMap featureMap)

Description copied from class: LibLinearTrainer

Extracts the features and Outputs in LibLinear's format.

Specified by:

extractData in class LibLinearTrainer<Regressor>

Parameters:

data - The input data.

outputInfo - The output info.

featureMap - The feature info.

Returns:

The features and outputs.