Module meta_transfer 

Structs

DomainAdaptation

Domain adaptation state: the characteristics of the source and target domains and the resulting per-feature adaptation weights.

TransferLearning

Transfer learning system: stores per-source-domain experiences and replays them into a target domain, weighted by measured domain similarity.

TransferMetrics

Measured transfer-learning outcomes.

Enums

TransferStrategy

Transfer learning strategies.

Constants

MIN_TRANSFER_SIMILARITY

Minimum cosine similarity between domains before source experiences are replayed into the target at all. Below this, transfer is more likely to be negative than helpful, so nothing is transferred.