Simulation of variational Gaussian process NARX models with GPGPU

Gaussian processes (GP) regression is a powerful probabilistic tool for modeling nonlinear dynamical systems. The downside of the method its cubic computational complexity with respect to training data that can be partially reduced using pseudo-inputs. dynamics represented an autoregressive model, which simplifies static case. When simulating uncertainty propagated through function and simulation cannot evaluated in closed-form. This paper combines variational methods GP approximations model exogenous inputs (NARX) form (VGP-NARX) models. We show how VGP-NARX models, on average, better approximate full GP-NARX than more commonly used (FITC) 10 chaotic time-series. capabilities models are compared existing approaches two benchmarks advantage general-purpose computing graphics processing units (GPGPU) Monte Carlo large validation sets addressed.