org.deeplearning4j.optimize.solvers.accumulation.encoding

Interface ThresholdAlgorithmReducer

All Superinterfaces:

Serializable

All Known Implementing Classes:

FixedThresholdAlgorithm.FixedAlgorithmThresholdReducer

public interface ThresholdAlgorithmReducer
extends Serializable

Used to combine ThresholdAlgorithm implementations in a way that is useful in a distributed training setting. ThresholdAlgorithm instances can be combined to save their historical state (if any) that is used to make decisions regarding the encoding threshold to use. Without combining and storing the state, any threshold adaption would be lost between epochs.

Author:

Alex Black

Method Summary

Modifier and Type	Method and Description
void	add(ThresholdAlgorithm instance) Add a ThresholdAlgorithm instance to the reducer
ThresholdAlgorithm	getFinalResult()
ThresholdAlgorithmReducer	merge(ThresholdAlgorithmReducer other) Combine two reducers and return the result

Method Detail

add

void add(ThresholdAlgorithm instance)

Add a ThresholdAlgorithm instance to the reducer

Parameters:

instance - Instance to add. May be null.

merge

ThresholdAlgorithmReducer merge(ThresholdAlgorithmReducer other)

Combine two reducers and return the result

Parameters:

other - Other reducer to combine with this one

Returns:

Combined reducer

getFinalResult

ThresholdAlgorithm getFinalResult()

Returns:

The final threshold reducer after combining all instances