Unlock: Meta-Learning

Learning to learn: find model initializations or embedding spaces that enable fast adaptation to new tasks from few examples. MAML, prototypical networks, and the connection to few-shot learning and in-context learning in LLMs.