Sparse Intrinsic Gaussian Processes for Prediction on Manifolds: Extending Applications to Environmental Contexts

Traditional Gaussian Processes are limited in their application by complex boundaries and intricately structured manifolds, such as when predicting water quality in the Aral Sea. Intrinsic Gaussian Processes adequately accommodate these complex conditions. To address the computational complexity of Intrinsic Gaussian Processes, we employ the sparse approximation method known as Deterministic Inducing Conditionals (DIC). The Sparse Intrinsic Gaussian Processes approach we propose offers effective prediction over such manifolds.