Active Learning

96 public types in this namespace, organized by kind.

Models & Types (50)

Type Summary
ActiveLearner<T, TInput, TOutput> Core implementation of the active learner that orchestrates the active learning loop.
ActiveLearningCandidate<T> One pool sample’s place in an active-learning ranking.
ActiveLearningContext<T> Context information for stopping criterion evaluation.
ActiveLearningIterationResult<T> Result from a single active learning iteration.
ActiveLearningResult<T> Final result from the complete active learning process.
ActiveLearningSelection<T> The result of an active-learning pass: a full ranking of the unlabeled pool plus the diversity-aware batch chosen for labeling.
BADGEStrategy<T, TInput, TOutput> BADGE (Batch Active learning by Diverse Gradient Embeddings) strategy.
BALDStrategy<T, TInput, TOutput> Bayesian Active Learning by Disagreement (BALD) strategy.
BALD<T> Implements Bayesian Active Learning by Disagreement (BALD) for sample selection.
BatchActiveLearner<T> Turns a per-sample informativeness (uncertainty) ranking into a diversity-aware batch selection, so the chosen batch covers the pool instead of collecting near-duplicate uncertain samples.
BatchBALD<T> Implements BatchBALD for joint batch selection in active learning.
BudgetExhaustedCriterion<T> Stopping criterion based on labeling budget exhaustion.
ClusteredBatchStrategy<T, TInput, TOutput> Clustering-based batch selection strategy using k-means clustering.
ColdStartAnalysisResult<T> Cold start analysis result showing initial sample selection performance.
CompositeCriterion<T> Composite stopping criterion that combines multiple criteria.
ConvergenceCriterion<T> Stopping criterion based on learning curve convergence.
CoreSetSelection<T> Implements core-set selection using the k-center-greedy algorithm for active learning.
CoreSetStrategy<T, TInput, TOutput> CoreSet strategy for active learning using geometric diversity.
DensityWeightedSampling<T> Implements density-weighted sampling for active learning.
DiversitySampling<T> Implements Diversity Sampling for active learning.
EntropySampling<T> Implements entropy sampling for active learning sample selection.
EvaluationMetrics<T> Evaluation metrics from model evaluation.
EvaluationResult<T> Result of evaluating a model.
ExpectedModelChange<T> Implements Expected Model Change (EMC) for active learning.
GradientBatchStrategy<T, TInput, TOutput> Gradient-based batch selection strategy using gradient embeddings (BADGE-style).
HybridSampling<T> Implements Hybrid Sampling that combines multiple active learning strategies.
InMemoryDataset<T, TInput, TOutput> In-memory implementation of a dataset for active learning.
InformationDensity<T> Implements information density sampling for active learning.
LearningCurve<T> Represents a learning curve showing performance vs.
LeastConfidenceSampling<T> Implements least confidence sampling for active learning sample selection.
MarginSampling<T> Implements margin sampling for active learning sample selection.
PerformancePlateauCriterion<T> Stopping criterion based on performance plateau detection.
PredictionStabilityCriterion<T> Stopping criterion based on prediction stability across iterations.
QueryByCommitteeStrategy<T, TInput, TOutput> Query By Committee (QBC) strategy for active learning.
QueryByCommittee<T> Implements Query-by-Committee (QBC) for active learning.
QueryQualityCriterion<T> Stopping criterion based on quality of remaining query candidates.
QueryStrategyMetrics<T> Detailed metrics for query strategy performance analysis.
RandomSampling<T> Implements random sampling for active learning (baseline strategy).
RankedBatchStrategy<T, TInput, TOutput> Simple ranked batch selection strategy with diversity filtering.
SamplesSelectedEventArgs<T, TInput> Event arguments for when samples are selected for labeling.
StrategyComparisonResult<T> Comparison result between different query strategies.
SubmodularBatchStrategy<T, TInput, TOutput> Submodular batch selection strategy using facility location objectives.
TimeBudgetCriterion<T> Stopping criterion based on time budget exhaustion.
TrainingMetrics<T> Training metrics from a model training iteration.
TrainingResult<T> Result of training a model.
UncertaintySamplingStrategy<T, TInput, TOutput> Uncertainty sampling strategy for active learning.
UncertaintySampling<T> Implements uncertainty sampling for active learning.
UncertaintyThresholdCriterion<T> Stopping criterion based on model uncertainty reaching a threshold.
UnlabeledPoolExhaustedCriterion<T> Stopping criterion that triggers when the unlabeled pool is exhausted.
VariationRatios<T> Implements variation ratios for active learning sample selection.

Base Classes (1)

Type Summary
DiversityStrategyBase<T, TInput, TOutput> Base class for diversity strategies with common functionality.

Interfaces (22)

Type Summary
IActiveLearner<T, TInput, TOutput> Interface for active learners.
IBatchStrategy<T, TInput, TOutput> Interface for batch selection strategies in active learning.
IBayesianStrategy<T, TInput, TOutput> Interface for Bayesian query strategies (e.g., BALD).
IClusteringBatchStrategy<T, TInput, TOutput> Interface for clustering-based batch selection.
ICommitteeStrategy<T, TInput, TOutput> Interface for committee-based query strategies (Query By Committee).
ICompositeCriterion<T> Interface for composite stopping criteria (multiple criteria combined).
IDensityWeightedStrategy<T, TInput, TOutput> Interface for density-weighted query strategies.
IDiversityStrategy<T, TInput, TOutput> Interface for diversity-based sampling strategies in active learning.
IDropoutModel<T, TInput, TOutput> Interface for models that support dropout inference.
IFeatureExtractor<T, TInput> Interface for models that can extract feature representations from inputs.
IFeatureExtractor<T, TInput> Interface for models that can extract feature representations.
IGradientBatchStrategy<T, TInput, TOutput> Interface for gradient-based batch selection (e.g., BADGE).
IGradientModel<T, TInput, TOutput> Interface for models that can compute gradient embeddings.
IOracle<TInput, TOutput> Interface for oracles (labeling providers) in active learning.
IPredictionBasedCriterion<T, TInput, TOutput> Interface for stopping criteria that need prediction access.
IProbabilisticModel<T, TInput, TOutput> Interface for models that can provide probability distributions.
IQueryStrategy<T, TInput, TOutput> Interface for query strategies in active learning.
IStoppingCriterion<T> Interface for stopping criteria in active learning.
ISubmodularBatchStrategy<T, TInput, TOutput> Interface for submodular batch selection strategies.
ITrainableModel<T, TInput, TOutput> Interface for models that support training on datasets.
IUncertaintyBasedCriterion<T> Interface for uncertainty-based stopping criteria.
IUncertaintyStrategy<T, TInput, TOutput> Interface for query strategies that support uncertainty-based selection.

Enums (21)

Type Summary
ChangeMetric<T> Defines the change metric to use.
ClusteringMethod Clustering methods for diversity-based sampling.
ColdStartStrategy Strategies for cold-start sample selection.
CombinationMethod<T> Defines methods for combining strategy scores.
CombinationMode Mode for combining multiple stopping criteria.
CommitteeDisagreementMeasure Disagreement measures for Query By Committee.
DisagreementMeasure Methods for measuring disagreement in a committee of models.
DisagreementMeasure<T> Defines the disagreement measure to use.
DistanceMetric Distance metrics for diversity-based sampling.
DistanceMetric<T> Defines the distance metric to use for core-set selection.
DistanceMetric<T> Defines the distance metric to use.
DistanceMetric Distance metrics for diversity-based strategies.
DiversityMethod<T> Defines the diversity selection method.
GradientApproximation Methods for approximating gradient-based importance.
QueryStrategyType Types of query strategies for active learning.
SimilarityMeasure<T> Defines the similarity measure for computing information density.
StabilityMeasure Stability measurement methods for prediction stability criterion.
StoppingCriterionType Criteria for early stopping in active learning.
SubmodularObjective Types of submodular objective functions.
UncertaintyMeasure Methods for measuring uncertainty in predictions.
UncertaintyMeasure<T> Defines the uncertainty measure to use.

Options & Configuration (1)

Type Summary
ActiveLearnerConfig<T> Comprehensive configuration for active learning.

Helpers & Utilities (1)

Type Summary
InMemoryDatasetFactory<T, TInput, TOutput> Factory implementation for creating in-memory datasets.