What is meta-learning (‘learning to learn’) and how does MAML implement it?

Question

Machine Learning — Hard

What is meta-learning (‘learning to learn’) and how does MAML implement it?

Accepted Answer

Meta-learning involves training models to quickly adapt to new tasks with limited examples. MAML specifically achieves this by finding initial parameters that allow for fast adaptation through second-order gradients.