Module active_learning 

Active Learning strategies for label-efficient training.

Strategies

• Uncertainty Sampling: Query most uncertain samples
• Query-by-Committee: Query samples with highest disagreement
• Entropy Sampling: Query highest entropy predictions

Structs

CoreSetSelection

Core-Set selection for diversity-based active learning.

EntropySampling

Entropy sampling: query samples with highest entropy.

ExpectedModelChange

Expected Model Change strategy.

MarginSampling

Margin sampling: query samples with smallest margin between top 2.

QueryByCommittee

Query-by-Committee: query samples with highest disagreement.

RandomSampling

Random sampling baseline.

UncertaintySampling

Uncertainty sampling: query samples with lowest confidence.

Traits

QueryStrategy

Active learning query strategy trait.