ws.palladian.classification.liblinear

Class LibLinearLearner

java.lang.Object

ws.palladian.core.AbstractLearner<LibLinearModel>

ws.palladian.classification.liblinear.LibLinearLearner

All Implemented Interfaces:

Learner<LibLinearModel>

public final class LibLinearLearner
extends AbstractLearner<LibLinearModel>

LIBLINEAR, A Library for Large Linear Classification. Wrapper for liblinear-java. For a documentation about liblinear see here and here. In addition, to the pure LIBLINEAR classifier, this wrapper adds the following functionality: a) Numerical data can be normalized using a Normalizer, b) nominal data is transformed to a numerical representation, using dummy coding (see DummyVariableCreator for more information).

Author:

Philipp Katz

Constructor Summary

Constructors

Constructor and Description
LibLinearLearner() Create a new LibLinearLearner with 'L2-regularized logistic regression', a cost value of 1.0 for constraints violation, a value of 0.01 as stopping criterion, a bias term of one, and Z-Score normalization for features.
LibLinearLearner(Normalizer normalizer) Create a new LibLinearLearner with 'L2-regularized logistic regression', a cost value of 1.0 for constraints violation, a value of 0.01 as stopping criterion, and a bias term of one
LibLinearLearner(de.bwaldvogel.liblinear.Parameter parameter, double bias, Normalizer normalizer) Create a new LibLinearLearner with the specified Parameter for training.

Method Summary

Modifier and Type	Method and Description
String	toString()
LibLinearModel	train(Dataset dataset)

Methods inherited from class ws.palladian.core.AbstractLearner

train, train

Methods inherited from class java.lang.Object

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

Constructor Detail

LibLinearLearner

public LibLinearLearner(de.bwaldvogel.liblinear.Parameter parameter,
                        double bias,
                        Normalizer normalizer)

Create a new LibLinearLearner with the specified Parameter for training.

Parameters:

parameter - The parameter, not null.

bias - The value for the bias term, use a value < 0 to add no bias term.

normalizer - The normalizer to use, not null. Use a NoNormalizer to skip normalization.

LibLinearLearner

public LibLinearLearner(Normalizer normalizer)

Create a new LibLinearLearner with 'L2-regularized logistic regression', a cost value of 1.0 for constraints violation, a value of 0.01 as stopping criterion, and a bias term of one

Parameters:

normalizer - The normalizer to use, not null. Use a NoNormalizer to skip normalization.

LibLinearLearner

public LibLinearLearner()

Create a new LibLinearLearner with 'L2-regularized logistic regression', a cost value of 1.0 for constraints violation, a value of 0.01 as stopping criterion, a bias term of one, and Z-Score normalization for features.

Method Detail

train

public LibLinearModel train(Dataset dataset)

toString

public String toString()

Overrides:

toString in class AbstractLearner<LibLinearModel>