org.deeplearning4j.optimize.solvers.accumulation

Class EncodedGradientsAccumulator.Builder

java.lang.Object

org.deeplearning4j.optimize.solvers.accumulation.EncodedGradientsAccumulator.Builder

Enclosing class:

EncodedGradientsAccumulator

public static class EncodedGradientsAccumulator.Builder
extends Object

Field Summary

Fields

Modifier and Type	Field and Description
protected Double	boundary
protected boolean	encodingDebugMode
protected MessageHandler	handler
protected long	initialMemory
protected int	parties
protected int	queueSize
protected ResidualPostProcessor	residualPostProcessor
protected ThresholdAlgorithm	thresholdAlgorithm

Constructor Summary

Constructors

Constructor and Description
Builder(int parties) This

Method Summary

Modifier and Type	Method and Description
EncodedGradientsAccumulator	build()
EncodedGradientsAccumulator.Builder	encodingDebugMode(boolean enable)
EncodedGradientsAccumulator.Builder	memoryParameters(long initialMemory, int queueSize) This method allows to define buffer memory parameters for this GradientsAccumulator Default values: 100MB initialMemory, 5 queueSize
EncodedGradientsAccumulator.Builder	messageHandler(MessageHandler handler) This method allows to specify MessageHandler instance Default value: EncodingHandler
EncodedGradientsAccumulator.Builder	residualPostProcessor(ResidualPostProcessor residualPostProcessor) Set the residual post processor
EncodedGradientsAccumulator.Builder	thresholdAlgorithm(ThresholdAlgorithm thresholdAlgorithm) This method allows to set the ThresholdAlgorithm to be used for determining the threshold
EncodedGradientsAccumulator.Builder	updatesBoundary(double boundary) This method enables optional limit for max number of updates per message Default value: 1.0 (no limit)

Methods inherited from class java.lang.Object

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

Field Detail

parties

protected int parties

thresholdAlgorithm

protected ThresholdAlgorithm thresholdAlgorithm

residualPostProcessor

protected ResidualPostProcessor residualPostProcessor

initialMemory

protected long initialMemory

queueSize

protected int queueSize

handler

protected MessageHandler handler

boundary

protected Double boundary

encodingDebugMode

protected boolean encodingDebugMode

Constructor Detail

Builder

public Builder(int parties)

This

Parameters:

parties -

Method Detail

messageHandler

public EncodedGradientsAccumulator.Builder messageHandler(@NonNull
                                                          MessageHandler handler)

This method allows to specify MessageHandler instance Default value: EncodingHandler

Parameters:

handler -

Returns:

thresholdAlgorithm

public EncodedGradientsAccumulator.Builder thresholdAlgorithm(ThresholdAlgorithm thresholdAlgorithm)

This method allows to set the ThresholdAlgorithm to be used for determining the threshold

Returns:

residualPostProcessor

public EncodedGradientsAccumulator.Builder residualPostProcessor(ResidualPostProcessor residualPostProcessor)

Set the residual post processor

updatesBoundary

public EncodedGradientsAccumulator.Builder updatesBoundary(double boundary)

This method enables optional limit for max number of updates per message Default value: 1.0 (no limit)

Parameters:

boundary - positive value in range 0..1

Returns:

memoryParameters

public EncodedGradientsAccumulator.Builder memoryParameters(long initialMemory,
                                                            int queueSize)

This method allows to define buffer memory parameters for this GradientsAccumulator Default values: 100MB initialMemory, 5 queueSize

Parameters:

initialMemory -

queueSize -

Returns:

encodingDebugMode

public EncodedGradientsAccumulator.Builder encodingDebugMode(boolean enable)

build

public EncodedGradientsAccumulator build()