Online Polynomial Chaos Kriging
This poster presents a stochastic meta-model that combines Polynomial Chaos Expansion (PCE) with Kriging while enabling online estimation as new data is available. PCE approximates the global behaviour, whereas Kriging model the latent correlation structure. The online adaptation is achieved by combining recursive least square and Bayesian state estimation techniques. The meta-model is illustrated over a non-static stochastic process benchmark, highlighting its capability to approximate the target and adapt in real time.
Palabras clave: Polynomial Chaos Expansion Kriging Gaussian Process Regression