org.deeplearning4j.earlystopping

Class EarlyStoppingConfiguration.Builder<T extends Model>

java.lang.Object

org.deeplearning4j.earlystopping.EarlyStoppingConfiguration.Builder<T>

Enclosing class:

EarlyStoppingConfiguration<T extends Model>

public static class EarlyStoppingConfiguration.Builder<T extends Model>
extends java.lang.Object

Constructor Summary

Constructors

Constructor and Description
Builder()

Method Summary

Modifier and Type	Method and Description
EarlyStoppingConfiguration<T>	build() Create the early stopping configuration
EarlyStoppingConfiguration.Builder<T>	epochTerminationConditions(EpochTerminationCondition... terminationConditions) Termination conditions to be evaluated every N epochs, with N set by evaluateEveryNEpochs option
EarlyStoppingConfiguration.Builder<T>	epochTerminationConditions(java.util.List<EpochTerminationCondition> terminationConditions) Termination conditions to be evaluated every N epochs, with N set by evaluateEveryNEpochs option
EarlyStoppingConfiguration.Builder<T>	evaluateEveryNEpochs(int everyNEpochs) How frequently should evaluations be conducted (in terms of epochs)? Defaults to every (1) epochs.
EarlyStoppingConfiguration.Builder<T>	iterationTerminationConditions(IterationTerminationCondition... terminationConditions) Termination conditions to be evaluated every iteration (minibatch)
EarlyStoppingConfiguration.Builder<T>	modelSaver(EarlyStoppingModelSaver<T> modelSaver) How should models be saved? (Default: in memory)
EarlyStoppingConfiguration.Builder<T>	saveLastModel(boolean saveLastModel) Save the last model? If true: save the most recent model at each epoch, in addition to the best model (whenever the best model improves).
EarlyStoppingConfiguration.Builder<T>	scoreCalculator(ScoreCalculator<T> scoreCalculator) Score calculator.

Methods inherited from class java.lang.Object

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

Constructor Detail

Builder

public Builder()

Method Detail

modelSaver

public EarlyStoppingConfiguration.Builder<T> modelSaver(EarlyStoppingModelSaver<T> modelSaver)

How should models be saved? (Default: in memory)

epochTerminationConditions

public EarlyStoppingConfiguration.Builder<T> epochTerminationConditions(EpochTerminationCondition... terminationConditions)

Termination conditions to be evaluated every N epochs, with N set by evaluateEveryNEpochs option

epochTerminationConditions

public EarlyStoppingConfiguration.Builder<T> epochTerminationConditions(java.util.List<EpochTerminationCondition> terminationConditions)

Termination conditions to be evaluated every N epochs, with N set by evaluateEveryNEpochs option

iterationTerminationConditions

public EarlyStoppingConfiguration.Builder<T> iterationTerminationConditions(IterationTerminationCondition... terminationConditions)

Termination conditions to be evaluated every iteration (minibatch)

saveLastModel

public EarlyStoppingConfiguration.Builder<T> saveLastModel(boolean saveLastModel)

Save the last model? If true: save the most recent model at each epoch, in addition to the best model (whenever the best model improves). If false: only save the best model. Default: false Useful for example if you might want to continue training after a max-time terminatino condition occurs.

evaluateEveryNEpochs

public EarlyStoppingConfiguration.Builder<T> evaluateEveryNEpochs(int everyNEpochs)

How frequently should evaluations be conducted (in terms of epochs)? Defaults to every (1) epochs.

scoreCalculator

public EarlyStoppingConfiguration.Builder<T> scoreCalculator(ScoreCalculator<T> scoreCalculator)

Score calculator. Used to calculate a score (such as loss function on a test set), every N epochs, where N is set by evaluateEveryNEpochs

build

public EarlyStoppingConfiguration<T> build()

Create the early stopping configuration