Uses of Interface
org.deeplearning4j.optimize.solvers.accumulation.encoding.ResidualPostProcessor

Packages that use ResidualPostProcessor

Package	Description
org.deeplearning4j.optimize.solvers.accumulation
org.deeplearning4j.optimize.solvers.accumulation.encoding
org.deeplearning4j.optimize.solvers.accumulation.encoding.residual

Uses of ResidualPostProcessor in org.deeplearning4j.optimize.solvers.accumulation

Fields in org.deeplearning4j.optimize.solvers.accumulation declared as ResidualPostProcessor

Modifier and Type	Field and Description
protected ResidualPostProcessor	EncodingHandler.initialResidualPostProcessor
protected ResidualPostProcessor	EncodedGradientsAccumulator.Builder.residualPostProcessor

Fields in org.deeplearning4j.optimize.solvers.accumulation with type parameters of type ResidualPostProcessor

Modifier and Type	Field and Description
protected ThreadLocal<ResidualPostProcessor>	EncodingHandler.residualPostProcessor

Methods in org.deeplearning4j.optimize.solvers.accumulation with parameters of type ResidualPostProcessor

Modifier and Type	Method and Description
EncodedGradientsAccumulator.Builder	EncodedGradientsAccumulator.Builder.residualPostProcessor(ResidualPostProcessor residualPostProcessor) Set the residual post processor

Constructors in org.deeplearning4j.optimize.solvers.accumulation with parameters of type ResidualPostProcessor

Constructor and Description
EncodedGradientsAccumulator(int parties, ThresholdAlgorithm thresholdAlgorithm, ResidualPostProcessor residualPostProcessor, boolean encodingDebugMode)
EncodingHandler(ThresholdAlgorithm thresholdAlgorithm, ResidualPostProcessor residualPostProcessor, Integer boundary, boolean encodingDebugMode)

Uses of ResidualPostProcessor in org.deeplearning4j.optimize.solvers.accumulation.encoding

Methods in org.deeplearning4j.optimize.solvers.accumulation.encoding that return ResidualPostProcessor

Modifier and Type	Method and Description
ResidualPostProcessor	ResidualPostProcessor.clone()

Uses of ResidualPostProcessor in org.deeplearning4j.optimize.solvers.accumulation.encoding.residual

Classes in org.deeplearning4j.optimize.solvers.accumulation.encoding.residual that implement ResidualPostProcessor

Modifier and Type	Class and Description
class	NoOpResidualPostProcessor This residual post process is a "no op" post processor.
class	ResidualClippingPostProcessor Residual clipping post processor clips the values of a residual every N iterations as follows: For residual vector R, and C = thresholdMultipleClipValue, T is the current encoding threshold R[i] = C*T if R[i] > C*T R[i] = -C*T if R[i] < -C*T R[i] is unmodified otherwise Note: Regarding the frequency, a value around 5 is suggested as a good balance between applying frequently enough, and minimizing the computational overhead.

Methods in org.deeplearning4j.optimize.solvers.accumulation.encoding.residual that return ResidualPostProcessor

Modifier and Type	Method and Description
ResidualPostProcessor	NoOpResidualPostProcessor.clone()