Package org.deeplearning4j.optimize.solvers.accumulation

Interface GradientsAccumulator

All Superinterfaces:

Serializable

All Known Implementing Classes:

BasicGradientsAccumulator, EncodedGradientsAccumulator

public interface GradientsAccumulator
extends Serializable

Method Summary

Modifier and Type	Method	Description
void	applyUpdate(StepFunction function, INDArray params, INDArray updates, boolean isFinalStep)	This method applies accumulated updates via given StepFunction
void	applyUpdate(StepFunction function, INDArray params, INDArray updates, double alpha)	This method applies accumulated updates via given StepFunction
IndexedTail	getExternalSource()
boolean	hasAnything()	This method checks if there are any (probably external) updates available
void	markExternalUpdates(boolean updatesAvailable)	This method allows to highlight early availability of updates
void	receiveUpdate(INDArray array)	This method accepts updates suitable for StepFunction and puts them to the queue, which is used in backpropagation loop PLEASE NOTE: array is expected to be ready for use and match params dimensionality
void	reset()	This method resets all accumulated updates (if any)
void	setExternalSource(IndexedTail source)	This method allows to pass external updates to accumulator, they will be populated across all workers using this GradientsAccumulator instance
void	storeUpdate(INDArray array, int iterationNumber, int epochNumber)	This method accepts updates suitable for StepFunction, and accumulates/propagates it across all workers
void	touch()	This method does initialization of given worker wrt Thread-Device Affinity

Method Detail

setExternalSource

void setExternalSource(IndexedTail source)

This method allows to pass external updates to accumulator, they will be populated across all workers using this GradientsAccumulator instance

Parameters:

source -

getExternalSource

IndexedTail getExternalSource()

applyUpdate

void applyUpdate(StepFunction function,
                 INDArray params,
                 INDArray updates,
                 boolean isFinalStep)

This method applies accumulated updates via given StepFunction

Parameters:

function -

params -

applyUpdate

void applyUpdate(StepFunction function,
                 INDArray params,
                 INDArray updates,
                 double alpha)

This method applies accumulated updates via given StepFunction

Parameters:

function -

params -

storeUpdate

void storeUpdate(INDArray array,
                 int iterationNumber,
                 int epochNumber)

This method accepts updates suitable for StepFunction, and accumulates/propagates it across all workers

Parameters:

array -

receiveUpdate

void receiveUpdate(INDArray array)

This method accepts updates suitable for StepFunction and puts them to the queue, which is used in backpropagation loop PLEASE NOTE: array is expected to be ready for use and match params dimensionality

Parameters:

array -

markExternalUpdates

void markExternalUpdates(boolean updatesAvailable)

This method allows to highlight early availability of updates

Parameters:

updatesAvailable -

reset

void reset()

This method resets all accumulated updates (if any)

touch

void touch()

This method does initialization of given worker wrt Thread-Device Affinity

hasAnything

boolean hasAnything()

This method checks if there are any (probably external) updates available

Returns: