org.encog.neural.networks.training.cross
Class CrossValidationKFold

java.lang.Object
  org.encog.neural.networks.training.BasicTraining
      org.encog.neural.networks.training.cross.CrossTraining
          org.encog.neural.networks.training.cross.CrossValidationKFold

All Implemented Interfaces:

Train

public class CrossValidationKFold
extends CrossTraining

Train using K-Fold cross validation. Each iteration will train a number of times equal to the number of folds - 1. Each of these sub iterations will train all of the data minus the fold. The fold is used to validate. Therefore, you are seeing an error that reflects data that was not always used as part of training. This should give you a better error result based on how the network will perform on non-trained data.(validation). The cross validation trainer must be provided with some other sort of trainer, perhaps RPROP, to actually perform the training. The training data must be the FoldedDataSet. The folded dataset can wrap most other training sets.

Constructor Summary

CrossValidationKFold(Train train, int k)
Construct a cross validation trainer.

Method Summary

void	iteration() Perform one iteration.

Methods inherited from class org.encog.neural.networks.training.cross.CrossTraining

getFolded, getNetwork

Methods inherited from class org.encog.neural.networks.training.BasicTraining

addStrategy, finishTraining, getCloud, getError, getIteration, getStrategies, getTraining, isTrainingDone, iteration, postIteration, preIteration, setCloud, setError, setIteration, setTraining

Methods inherited from class java.lang.Object

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

Constructor Detail

CrossValidationKFold

public CrossValidationKFold(Train train,
                            int k)

Construct a cross validation trainer.

Parameters:

train - The training

k - The number of folds.

Method Detail

iteration

public void iteration()

Perform one iteration.