Bayesian Nonparametric Mixed-Effect ODEs with Gaussian Processes
MEGPODE is a Bayesian nonparametric model for differential equations that decomposes subject dynamics into a shared population component and individual deviations using Gaussian Processes. It translates classical population-level effects into function space, allowing flexible personalization of dynamics. The method uses state-space GPs and collocation constraints to jointly infer both the trajectories and the underlying vector fields.