New paper published in in Computational Methods in Applied Mathematics and Engineering

In the paper entitled A spectral surrogate model for stochastic simulators computed from trajectory samples, we develop a new surrogate modeling method for a type of stochastic simulators for which we can sample from trajectories, i.e., obtain samples while keeping the underlying stochasticity fixed. We propose to use polynomial chaos expansion (PCE) to approximate the trajectories and subsequently apply extended Karhunen-Loève expansion (KLE) in the weighted L2-space to reduce the dimensionality of the stochastic space. The joint distribution of the random coefficients of the KLE is inferred using the marginal-vine copula framework. The resulting surrogate model approximates mean and variance of the true model and can generate new trajectories.

For more information, please follow this external page link or the publication and this link for the associated report on our internal archive.