Scalable Exact Inference via Adaptive Multi-Resolution Gaussian Processes

This framework enables scalable, exact inference in Gaussian processes by constructing naturally data-sparse covariance matrices using adaptive multi-resolution basis functions. By anchoring basis functions directly to samples, it eliminates auxiliary points and uses sparse Cholesky inversion for efficiency.