ISubmodularBatchStrategy<T, TInput, TOutput>

Interfaces · AiDotNet.ActiveLearning.Interfaces

Interface for submodular batch selection strategies.

Implements: IBatchStrategy<T, TInput, TOutput>

Known implementations: SubmodularBatchStrategy<T, TInput, TOutput>

For Beginners

Submodular functions have a “diminishing returns” property: adding more similar samples provides less benefit. This naturally encourages diversity.

Methods

Method Summary
T ComputeMarginalGain(IReadOnlyList<int> currentSelection, int candidateIndex, IDataset<T, TInput, TOutput> unlabeledPool) Computes the marginal gain of adding a sample to the current selection.
int[] GreedyMaximization(int[] candidateIndices, IDataset<T, TInput, TOutput> unlabeledPool, int batchSize) Performs greedy submodular maximization to select a batch.