Learning to (Learn at Test Time)

Yu Sun, Xinhao Li, Karan Dalal +6

We reformulate the problem of supervised learning as learning to learn with two nested loops (i.e. learning problems). The inner loop learns on each individual instance with self-s…