NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:9224
Title:Learning Data Manipulation for Augmentation and Weighting


		
The paper presents a gradient based meta learning approach to automating data augmentation and weighting examples. The experiments support the advantages of the proposed technique. There has been recent work on using gradient based meta-learning for weighting examples (Ren et al., ICML, 2018 [26]) and learning reward functions (Agarwal et al., ICML 2019 [https://arxiv.org/pdf/1902.07198.pdf]). There are some interesting technical novelty in the proposed algorithm and a clear discussion of such novelty in the context of recent papers is beneficial. Given the similarity of the proposed technique and existing recent work on meta-learning, I recommend accepting as a poster.