Meta-Learning and Universality: Deep Representations and Gradient Descent can Approximate any Learning Algorithm

[TMats]

https://arxiv.org/abs/1710.11622

• Chelsea Finn, Sergey Levine
• Submitted on 31 Oct 2017 (v1), last revised 14 Feb 2018 (this version, v3)