Module meta_learner 

MetaLearner — MAML-inspired meta-learning system.

Maintains task-specific adaptations and a shared meta-representation. The outer loop (meta-update) aggregates task adaptations to improve the shared initialisation; the inner loop (adapt_to_task) fine-tunes the shared weights for a single task in a small number of gradient-descent steps.

Structs

MetaLearner

MAML-inspired meta-learner.

MetaLearnerConfig

Configuration for MetaLearner.

MetaLearnerStats

Aggregate statistics for a MetaLearner instance.

MetaParameters

Shared meta-parameters: a linear model with dims weights and one bias.

MetaTask

A meta-learning task consisting of a support set (used for inner-loop adaptation) and a query set (used to evaluate the adapted model).

TaskAdaptation

The result of adapting the meta-parameters to a specific task.

TaskExample

A single labeled training example.

TaskId

Newtype wrapper around a task identifier string.

Enums

MetaError

Errors produced by MetaLearner operations.

TaskType

Discriminates the learning objective of a meta-task.