Full scale multi-output Gaussian process emulator with nonseparable auto-covariance functions

摘要: Gaussian process emulator with separable covariance function has been utilized extensively in modeling large computer model outputs. The assumption of separability imposes constraints on the and may negatively affect its performance some applications where not hold. We propose a multi-output nonseparable auto-covariance to avoid limitations using emulators. In addition, facilitate computation emulator, we introduce new computational method, referred as Full-Scale approximation method block modulating (FSA-Block) approach. FSA-Block is an effective accurate reduce computations for models, which applies both partially models. illustrate effectiveness our through simulation studies compare it emulators covariances. also apply real code carbon capture system.