Meta-learning

Meta-learning trains models to learn new tasks quickly by learning across tasks, often via gradient-based or metric-based approaches. Popular methods include MAML, prototypical networks, and optimization-based meta-learners for few-shot adaptation.