org.apache.ignite.ml.optimization.updatecalculators

Class SimpleGDUpdateCalculator

java.lang.Object

org.apache.ignite.ml.optimization.updatecalculators.SimpleGDUpdateCalculator

All Implemented Interfaces:

Serializable, ParameterUpdateCalculator<SmoothParametrized,SimpleGDParameterUpdate>

public class SimpleGDUpdateCalculator
extends Object
implements ParameterUpdateCalculator<SmoothParametrized,SimpleGDParameterUpdate>

Simple gradient descent parameters updater.

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field and Description
protected IgniteFunction<Vector,IgniteDifferentiableVectorToDoubleFunction>	loss Loss function.

Constructor Summary

Constructors

Constructor and Description
SimpleGDUpdateCalculator() Construct instance of this class with default parameters.
SimpleGDUpdateCalculator(double learningRate) Construct SimpleGDUpdateCalculator.

Method Summary

Modifier and Type	Method and Description
SimpleGDParameterUpdate	calculateNewUpdate(SmoothParametrized mlp, SimpleGDParameterUpdate updaterParameters, int iteration, Matrix inputs, Matrix groundTruth) Calculate new update.
SimpleGDParameterUpdate	init(SmoothParametrized mdl, IgniteFunction<Vector,IgniteDifferentiableVectorToDoubleFunction> loss) Initializes the update calculator.
<M1 extends SmoothParametrized> M1	update(M1 obj, SimpleGDParameterUpdate update) Update given obj with this parameters.
SimpleGDUpdateCalculator	withLearningRate(double learningRate) Create new instance of this class with same parameters as this one, but with new learning rate.

Methods inherited from class java.lang.Object

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

Field Detail

loss

protected IgniteFunction<Vector,IgniteDifferentiableVectorToDoubleFunction> loss

Loss function.

Constructor Detail

SimpleGDUpdateCalculator

public SimpleGDUpdateCalculator()

Construct instance of this class with default parameters.

SimpleGDUpdateCalculator

public SimpleGDUpdateCalculator(double learningRate)

Construct SimpleGDUpdateCalculator.

Parameters:

learningRate - Learning rate.

Method Detail

init

public SimpleGDParameterUpdate init(SmoothParametrized mdl,
                                    IgniteFunction<Vector,IgniteDifferentiableVectorToDoubleFunction> loss)

Initializes the update calculator.

Specified by:

init in interface ParameterUpdateCalculator<SmoothParametrized,SimpleGDParameterUpdate>

Parameters:

mdl - Model to be trained.

loss - Loss function.

Returns:

Initialized parameters.

calculateNewUpdate

public SimpleGDParameterUpdate calculateNewUpdate(SmoothParametrized mlp,
                                                  SimpleGDParameterUpdate updaterParameters,
                                                  int iteration,
                                                  Matrix inputs,
                                                  Matrix groundTruth)

Calculate new update.

Specified by:

calculateNewUpdate in interface ParameterUpdateCalculator<SmoothParametrized,SimpleGDParameterUpdate>

Parameters:

mlp - Model to be updated.

updaterParameters - Updater parameters to updateCache.

iteration - Current trainer iteration.

inputs - Inputs.

groundTruth - True values.

Returns:

Updated parameters.

update

public <M1 extends SmoothParametrized> M1 update(M1 obj,
                                                 SimpleGDParameterUpdate update)

Update given obj with this parameters.

Specified by:

update in interface ParameterUpdateCalculator<SmoothParametrized,SimpleGDParameterUpdate>

Parameters:

obj - Object to be updated.

withLearningRate

public SimpleGDUpdateCalculator withLearningRate(double learningRate)

Create new instance of this class with same parameters as this one, but with new learning rate.

Parameters:

learningRate - Learning rate.

Returns:

New instance of this class with same parameters as this one, but with new learning rate.