QueryByCommittee<T>

Models & Types · AiDotNet.ActiveLearning

Implements Query-by-Committee (QBC) for active learning.

Implements: IActiveLearningStrategy<T>

For Beginners

Query-by-Committee uses multiple models (a “committee”) to evaluate samples. It selects samples where the committee members disagree the most. The intuition is that disagreement indicates uncertainty in the version space - the region of hypotheses consistent with the labeled data.

How It Works

How it works:

Reference: Seung et al., “Query by Committee” (1992). COLT.

Constructors

Constructor Summary
QueryByCommittee(IEnumerable<IFullModel<T, Tensor<T>, Tensor<T>>> committee, DisagreementMeasure<T> measure) Initializes a new instance of the QueryByCommittee class.

Properties

Property Summary
IReadOnlyList<IFullModel<T, Tensor<T>, Tensor<T>>> Committee { get; } Gets the committee of models.
bool UseBatchDiversity { get; set; }
string Name { get; }

Methods

Method Summary
Dictionary<string, T> GetSelectionStatistics()
Vector<T> ComputeInformativenessScores(IFullModel<T, Tensor<T>, Tensor<T>> model, Tensor<T> unlabeledPool)
int[] SelectSamples(IFullModel<T, Tensor<T>, Tensor<T>> model, Tensor<T> unlabeledPool, int batchSize)