Package org.tribuo.common.sgd

Class AbstractLinearSGDTrainer<T extends Output<T>,U,V extends AbstractLinearSGDModel<T>>

java.lang.Object

org.tribuo.common.sgd.AbstractSGDTrainer<T,U,V,LinearParameters>

org.tribuo.common.sgd.AbstractLinearSGDTrainer<T,U,V>

All Implemented Interfaces:

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

Direct Known Subclasses:

LinearSGDTrainer, LinearSGDTrainer, LinearSGDTrainer

public abstract class AbstractLinearSGDTrainer<T extends Output<T>,U,V extends AbstractLinearSGDModel<T>>
extends AbstractSGDTrainer<T,U,V,LinearParameters>

A trainer for a linear model which uses SGD.

It's an AbstractSGDTrainer operating on LinearParameters, with the bias folded into the features.

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, INCREMENT_INVOCATION_COUNT

Constructor Summary

Constructors

Modifier	Constructor	Description
protected	AbstractLinearSGDTrainer()	For olcut.
protected	AbstractLinearSGDTrainer(StochasticGradientOptimiser optimiser, int epochs, int loggingInterval, int minibatchSize, long seed)	Constructs an SGD trainer for a linear model.

Method Summary

Modifier and Type	Method	Description
protected LinearParameters	createParameters(int numFeatures, int numOutputs, SplittableRandom localRNG)	Constructs the trainable parameters object, in this case a LinearParameters containing a single weight matrix.
protected String	getName()	Returns the default model name.

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

createModel, getInvocationCount, getModelClassName, getObjective, getProvenance, getTarget, postConfig, setInvocationCount, setShuffle, shuffleInPlace, train, train, train

Methods inherited from class java.lang.Object

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

Constructor Details

AbstractLinearSGDTrainer

protected AbstractLinearSGDTrainer(StochasticGradientOptimiser optimiser,
 int epochs,
 int loggingInterval,
 int minibatchSize,
 long seed)

Constructs an SGD trainer for a linear model.

Parameters:

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.

AbstractLinearSGDTrainer

protected AbstractLinearSGDTrainer()

For olcut.

Method Details

getName

protected String getName()

Returns the default model name.

Specified by:

getName in class AbstractSGDTrainer<T extends Output<T>,U,V extends AbstractLinearSGDModel<T>,LinearParameters>

Returns:

The default model name.

createParameters

protected LinearParameters createParameters(int numFeatures,
 int numOutputs,
 SplittableRandom localRNG)

Constructs the trainable parameters object, in this case a LinearParameters containing a single weight matrix.

Specified by:

createParameters in class AbstractSGDTrainer<T extends Output<T>,U,V extends AbstractLinearSGDModel<T>,LinearParameters>

Parameters:

numFeatures - The number of input features.

numOutputs - The number of output dimensions.

localRNG - The RNG to use for parameter initialisation.

Returns:

The trainable parameters.