Tilting the playing field: Dynamical loss functions for machine learning

           Miguel Ruiz-Garcia 1 2 Ge Zhang 1 Samuel S. Schoenholz 3 Andrea J. Liu 1

             Abstract                 tion scale (Glorot & Bengio, 2010; Xiao et al., 2018) or
                                 learning rate schedule (He et al., 2016); the neural network
   We show that learning can be improved by us-         architecture itself (Zoph & Le, 2016); the types of data aug-
   ing loss functions that evolve  ...