Multiple shooting for training neural differential equations on time series

Evren Mert Turan, Johannes Jäschke

Neural differential equations have recently emerged as a flexible data-driven/hybrid approach to model time-series data. This work experimentally demonstrates that if the data cont…