The authors introduce a novel conservative machine learning-based semi-Lagrangian finite difference method that allows for extra-large time step evolution and significantly simplifies algorithm implementation compared to traditional numerical schemes.
