Enum ActiveLearningStrategyType

Namespace

AiDotNet.Models.Options

Assembly

AiDotNet.dll

Specifies the active learning strategy to use for sample selection.

public enum ActiveLearningStrategyType

Fields

BALD = 13

Bayesian Active Learning by Disagreement - uses mutual information between predictions and model parameters for selection.

BatchBALD = 14

Batch-mode BALD that accounts for redundancy when selecting multiple samples.

CoreSetSelection = 10

Selects representative samples that form a core set of the data.

DensityWeightedSampling = 9

Weights uncertainty by sample density in the input space.

DiversitySampling = 4

Selects samples that maximize coverage of the input space.

EntropySampling = 6

Selects samples based on entropy of predicted class probabilities.

ExpectedModelChange = 3

Selects samples that would cause the largest change to model parameters.

HybridSampling = 5

Combines multiple criteria (uncertainty + diversity) for selection.

InformationDensity = 11

Combines uncertainty with representativeness based on local density.

LeastConfidenceSampling = 8

Selects samples where the top prediction has the lowest confidence.

MarginSampling = 7

Selects samples with the smallest margin between top two predictions.

QueryByCommittee = 2

Uses multiple models and selects samples where they disagree the most.

Random = 0

Selects samples randomly. Simple baseline strategy.

UncertaintySampling = 1

Selects samples where the model is most uncertain about predictions. Supports multiple uncertainty measures (least confidence, margin, entropy).

VariationRatios = 12

Uses variation ratios (1 - max probability) for uncertainty estimation.

Remarks

For Beginners: Active learning strategies help identify which unlabeled samples would be most valuable to label. Different strategies use different criteria to measure how "informative" a sample is for training the model.