Meta-learning algorithms aim to learn two components: a model that predicts targets for a task, and a base learner that updates that model when given examples from a new task. This additional level of learning can be powerful, but it also creates another potential source of overfitting, since we can now overfit in either the model or the base learner.

I did not understand the experimental setup for the sinusoid experiment in Section 5.2 (lines 238-241). The 10 disjoint intervals do not cover the whole domain [-5, 5]. What happens if x lies in (-4.5, -4) for example? Line 241: "there exists a continuous function that covers each piecewise component."; An illustration of this task (possibly in the Appendix) would make things clearer.