NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:60
Title:Meta-Learning with Implicit Gradients


		
This paper elaborates MAML with implicit gradients, avoiding long inner-loop chains appearing in the MAML. The idea of incorporating implicit differentiation into MAML is interesting, without any doubt. In general, the paper is well written, although it is not clear what iMAML should give up. During the discussion period, two of reviewers were positive and even one reviewer raised his/her score. However, one reviewer declined to strongly support this paper. While the paper is not perfect, I believe that the paper is deserved to be presented at NeurIPS.