Attentive Neural Processes and Batch Bayesian Optimization for Scalable Calibration of Physics-Informed Digital Twins

Physics-informed dynamical system models form critical components of digital\ntwins of the built environment. These digital twins enable the design of\nenergy-efficient infrastructure, but must be properly calibrated to accurately\nreflect system behavior for downstream prediction and analysis. Dynamical\nsystem models of modern buildings are typically described by a large number of\nparameters and incur significant computational expenditure during simulations.\nTo handle large-scale calibration of digital twins without exorbitant\nsimulations, we propose ANP-BBO: a scalable and parallelizable batch-wise\nBayesian optimization (BBO) methodology that leverages attentive neural\nprocesses (ANPs).\n