NeurIPS-2019
means being highly related to my personal research interest.
Example Weighting by Meta Learning for Robustness
-
Meta-Weight-Net: Learning an Explicit Mapping For Sample Weighting
- A following work of https://xinshaoamoswang.github.io/paperlists/2018-12-30-ICML/#example-weighting-by-meta-learning-gradient-directions
- We propose to automatically learn an explicit loss-weight function, parameterized by an MLP from data in a meta-learning manner. Due to the universal approximation capability of this weight net, it can finely fit a wide range of weighting functions including those used in conventional research.
- The sample weights of those samples better complying with the meta-data knowledge will be improved, while those violating such meta-knowledge will be suppressed.