Nested Learning: A new ML paradigm for continual learning

(research.google)

6 points | by pr337h4m 13 hours ago ago

1 comments

  • in-silico 9 hours ago

    The ideas are very interesting, but the paper is lacking in details (especially given that they defer several times to an appendix which is not present in their pdf). I don't think this could be reproduced very closely given what they tell us.

    I feel like this paper's main contribution is introducing the concept of nested learning and getting people thinking about it, while a later paper that figures out the engineering will get more attention.