maml (model-agnostic meta-learning)

An approach for quickly learning new tasks, in a few steps, with the aim of using these learned algorithms for meta-learning.