DynGMA: A robust approach for learning stochastic differential equations from data

Aiqing Zhu, Qianxiao Li

Abstract: Highlights•Novel density approximations are proposed for learning SDE via maximum likelihood estimation.•The proposed density approximation allows for multiple time steps, effectively reducing discretization errors.•The proposed method exhibits superior accuracy in learning the governing functions and computing the invariant distribution.•The proposed method is capable of handling trajectory data with low time resolution and variable time step sizes.