IStoppingCriterion<T>

Interfaces · AiDotNet.ActiveLearning.Interfaces

Interface for stopping criteria in active learning.

Known implementations: BudgetExhaustedCriterion<T>, ConvergenceCriterion<T>, ICompositeCriterion<T>, IPredictionBasedCriterion<T, TInput, TOutput>, IUncertaintyBasedCriterion<T>, PerformancePlateauCriterion<T>, PredictionStabilityCriterion<T>, QueryQualityCriterion<T>, TimeBudgetCriterion<T>, UnlabeledPoolExhaustedCriterion<T>

For Beginners

Active learning loops can be stopped based on various criteria beyond just exhausting the labeling budget. Early stopping can save resources when additional labels won’t significantly improve the model.

How It Works

Common Stopping Criteria:

Properties

Property Summary
string Description { get; } Gets a description of when this criterion triggers.
string Name { get; } Gets the name of the stopping criterion.

Methods

Method Summary
T GetProgress(ActiveLearningContext<T> context) Gets a progress indicator (0 to 1) showing how close to stopping.
bool ShouldStop(ActiveLearningContext<T> context) Checks whether the stopping criterion is met.
void Reset() Resets the criterion to its initial state.