Neural Rough Differential Equations for Long Time Series

            James Morrill 1 2 Cristopher Salvi 1 2 Patrick Kidger 1 2 James Foster 1 2

             Abstract                  direct way to modify the trajectory given subsequent ob-
   Neural controlled differential equations (CDEs)         servations. In contrast, the vector field of a Neural CDE
   are the continuous-time analogue of recurrent neu-        depends upon the time-varying data, so that the trajectory
   ral networks, as N ...