Core ML Concepts · advanced · concept 32 of 176
Meta-Learning (Learning to Learn)
Training a model across many small tasks so it can pick up a brand-new task from a handful of examples. The explicit version of what LLMs turned out to do implicitly, in-context learning at scale made dedicated meta-learning algorithms like MAML mostly historical.
Key terms
MAMLFew-shot tasksInner and outer loopTask distributionIn-context learning
Learn these first
Where you meet it in the real world
Few-shot classification, rapid adaptation in robotics, historically the road to in-context learning
Videos
▶ Stanford CS330 Deep Multi-Task & Meta Learning - What is multi-task learning? I 2022 I Lecture 1 ↗
Stanford Online · YouTube
Guides and articles
2018-11-30-meta-learning ↗
Lil'Log (OpenAI researcher)