com.tibco.patterns.learn.training

Interface TrainingObserver

All Known Subinterfaces:

ConvergenceObserver

All Known Implementing Classes:

COErrorRate, COErrorRateMin, COIterationCount, COScoreChange

public interface TrainingObserver

Defines an object that gathers training statistics for a certain dataset.

Method Summary

Modifier and Type	Method and Description
void	beginIteration() Resets internal state to prepare for a new iteration of evaluating the entire dataset.
void	endIteration() Any actions that need to be performed after evaluating the dataset.
void	evaluatePrediction(VectorExample e) Interprets or accumulates prediction result of a single feature vector Called once for each feature vector in the dataset
boolean	hasIterations()
boolean	isPerfectResult()

Method Detail

beginIteration

void beginIteration()

Resets internal state to prepare for a new iteration of evaluating the entire dataset. Can also be used to report progress before every training iteration.

evaluatePrediction

void evaluatePrediction(VectorExample e)

Interprets or accumulates prediction result of a single feature vector Called once for each feature vector in the dataset

Parameters:

e - - an example that must have the RLink prediction assigned.

endIteration

void endIteration()

Any actions that need to be performed after evaluating the dataset. E.g. report progress in GUI.

isPerfectResult

boolean isPerfectResult()

Returns:

true if the model fits the training data perfectly and therefore its predictions cannot be improved. Any further training with the same data is futile.

hasIterations

boolean hasIterations()

Returns:

true if any training iterations have been done using this observer.