QueryStrategyType

Enums · AiDotNet.ActiveLearning.Config

Types of query strategies for active learning.

Fields

Field Summary
QueryStrategyType BADGE BADGE: Batch Active learning by Diverse Gradient Embeddings.
QueryStrategyType BALD Bayesian Active Learning by Disagreement - uses MC Dropout.
QueryStrategyType BatchBALD Batch variant of BALD that considers joint information.
QueryStrategyType CoreSet Select samples that are representative of the unlabeled pool.
QueryStrategyType Diversity Select diverse samples in feature space.
QueryStrategyType Entropy Select samples with highest prediction entropy.
QueryStrategyType ExpectedErrorReduction Select samples that would reduce expected error most.
QueryStrategyType ExpectedModelChange Select samples that would cause largest gradient update.
QueryStrategyType InformationDensity Combine uncertainty with density in feature space.
QueryStrategyType LearningLoss Learning Loss - learns to predict loss for query selection.
QueryStrategyType LeastConfidence Select samples with lowest maximum predicted probability.
QueryStrategyType Margin Select samples with smallest margin between top predictions.
QueryStrategyType QBC Query By Committee - uses ensemble disagreement.
QueryStrategyType Random Random sampling baseline.
QueryStrategyType UncertaintySampling Query samples where the model is most uncertain.
QueryStrategyType VarianceReduction Select samples that would reduce prediction variance most.
int value__