Model-Agnostic Meta-Learning for Fast Adaptation…

MAML asks whether training should optimize a model's answer before adaptation or the quality of the answer after a few task-specific gradient steps. Read it to see how support and query examples create a nested objective, why one initialization can move toward several related tasks, what the sinusoid and MiniImageNet results establish, and where task shift and second-order cost bound the claim.

Reading focus: How a task becomes one outer-loop training example through separate support and query data. Why the meta-objective changes the usefulness of a local gradient rather than merely lowering pre-adaptation loss. How to read the MiniImageNet, sinusoid, and first-order results without turning related-task adaptation into a universal capability.

ICML 2017. Chelsea Finn, Pieter Abbeel, and Sergey Levine. 35 min read, medium difficulty.