Opening the Blackbox: Accelerating Neural Differential Equations by
           Regularizing Internal Solver Heuristics

              Avik Pal 1 2 Yingbo Ma 2 Viral Shah 2 Christopher Rackauckas 2 3 4 5

              Abstract
   Democratization of machine learning requires ar-
   chitectures that automatically adapt to new prob-
   lems. Neural Differential Equations (NDEs) have
   emerged as a popular modeling framework by re-
   moving the need for ML practitioners to choose
   the number of  ...