Class LinearSGDTrainer

java.lang.Object

org.tribuo.common.sgd.AbstractSGDTrainer<Regressor, DenseVector, AbstractLinearSGDModel<Regressor>, LinearParameters>

org.tribuo.common.sgd.AbstractLinearSGDTrainer<Regressor, DenseVector>

org.tribuo.regression.sgd.linear.LinearSGDTrainer

All Implemented Interfaces:

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

public class LinearSGDTrainer
extends AbstractLinearSGDTrainer<Regressor, DenseVector>

A trainer for a linear regression model which uses SGD. Independently trains each output dimension, unless they are tied together in the optimiser.

See:

 Bottou L.
 "Large-Scale Machine Learning with Stochastic Gradient Descent"
 Proceedings of COMPSTAT, 2010.
 

Field Summary

Fields inherited from class org.tribuo.common.sgd.AbstractSGDTrainer

addBias, epochs, loggingInterval, minibatchSize, optimiser, rng, seed, shuffle

Fields inherited from interface org.tribuo.Trainer

DEFAULT_SEED

Constructor Summary

Constructors

Constructor	Description
LinearSGDTrainer(RegressionObjective objective, StochasticGradientOptimiser optimiser, int epochs, int loggingInterval, int minibatchSize, long seed)	Constructs an SGD trainer for a linear model.
LinearSGDTrainer(RegressionObjective objective, StochasticGradientOptimiser optimiser, int epochs, int loggingInterval, long seed)	Constructs an SGD trainer for a linear model.
LinearSGDTrainer(RegressionObjective objective, StochasticGradientOptimiser optimiser, int epochs, long seed)	Constructs an SGD trainer for a linear model.

Method Summary

Modifier and Type	Method	Description
protected LinearSGDModel	createModel(String name, ModelProvenance provenance, ImmutableFeatureMap featureMap, ImmutableOutputInfo<Regressor> outputInfo, LinearParameters parameters)	Creates the appropriate model subclass for this subclass of AbstractSGDTrainer.
protected String	getModelClassName()	Returns the class name of the model that's produced by this trainer.
protected SGDObjective<DenseVector>	getObjective()	Returns the objective used by this trainer.
protected DenseVector	getTarget(ImmutableOutputInfo<Regressor> outputInfo, Regressor output)	Extracts the appropriate training time representation from the supplied output.
String	toString()

Methods inherited from class org.tribuo.common.sgd.AbstractLinearSGDTrainer

createParameters, getName

Methods inherited from class org.tribuo.common.sgd.AbstractSGDTrainer

getInvocationCount, getProvenance, postConfig, setShuffle, shuffleInPlace, train, train

Methods inherited from class java.lang.Object

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

Constructor Details

LinearSGDTrainer

public LinearSGDTrainer(RegressionObjective objective,
 StochasticGradientOptimiser optimiser,
 int epochs,
 int loggingInterval,
 int minibatchSize,
 long seed)

Constructs an SGD trainer for a linear model.

Parameters:

objective - The objective function to optimise.

optimiser - The gradient optimiser to use.

epochs - The number of epochs (complete passes through the training data).

loggingInterval - Log the loss after this many iterations. If -1 don't log anything.

minibatchSize - The size of any minibatches.

seed - A seed for the random number generator, used to shuffle the examples before each epoch.

LinearSGDTrainer

public LinearSGDTrainer(RegressionObjective objective,
 StochasticGradientOptimiser optimiser,
 int epochs,
 int loggingInterval,
 long seed)

Constructs an SGD trainer for a linear model.

Sets the minibatch size to 1.

Parameters:

objective - The objective function to optimise.

optimiser - The gradient optimiser to use.

epochs - The number of epochs (complete passes through the training data).

loggingInterval - Log the loss after this many iterations. If -1 don't log anything.

seed - A seed for the random number generator, used to shuffle the examples before each epoch.

LinearSGDTrainer

public LinearSGDTrainer(RegressionObjective objective,
 StochasticGradientOptimiser optimiser,
 int epochs,
 long seed)

Constructs an SGD trainer for a linear model.

Sets the minibatch size to 1 and the logging interval to 1000.

Parameters:

objective - The objective function to optimise.

optimiser - The gradient optimiser to use.

epochs - The number of epochs (complete passes through the training data).

seed - A seed for the random number generator, used to shuffle the examples before each epoch.

Method Details

getTarget

protected DenseVector getTarget(ImmutableOutputInfo<Regressor> outputInfo,
 Regressor output)

Description copied from class: AbstractSGDTrainer

Extracts the appropriate training time representation from the supplied output.

Specified by:

getTarget in class AbstractSGDTrainer<Regressor, DenseVector, AbstractLinearSGDModel<Regressor>, LinearParameters>

Parameters:

outputInfo - The output info to use.

output - The output to extract.

Returns:

The training time representation of the output.

getObjective

protected SGDObjective<DenseVector> getObjective()

Description copied from class: AbstractSGDTrainer

Returns the objective used by this trainer.

Specified by:

getObjective in class AbstractSGDTrainer<Regressor, DenseVector, AbstractLinearSGDModel<Regressor>, LinearParameters>

Returns:

The SGDObjective used by this trainer.

createModel

protected LinearSGDModel createModel(String name,
 ModelProvenance provenance,
 ImmutableFeatureMap featureMap,
 ImmutableOutputInfo<Regressor> outputInfo,
 LinearParameters parameters)

Description copied from class: AbstractSGDTrainer

Creates the appropriate model subclass for this subclass of AbstractSGDTrainer.

Specified by:

createModel in class AbstractSGDTrainer<Regressor, DenseVector, AbstractLinearSGDModel<Regressor>, LinearParameters>

Parameters:

name - The model name.

provenance - The model provenance.

featureMap - The feature map.

outputInfo - The output info.

parameters - The model parameters.

Returns:

A new instance of the appropriate subclass of Model.

getModelClassName

protected String getModelClassName()

Description copied from class: AbstractSGDTrainer

Returns the class name of the model that's produced by this trainer.

Specified by:

getModelClassName in class AbstractSGDTrainer<Regressor, DenseVector, AbstractLinearSGDModel<Regressor>, LinearParameters>

Returns:

The model class name;

toString

public String toString()

Overrides:

toString in class Object