Spatial Bayesian Neural Networks: A Flexible Approach for Modeling Spatial Processes

Spatial Bayesian neural networks (SBNNs) are a new, flexible class of spatial process models that leverage the representational capacity of Bayesian neural networks. SBNNs can match the finite-dimensional distributions of a wide range of target spatial processes, including stationary and non-stationary Gaussian processes, as well as lognormal processes.