SKIing on Simplices: Kernel Interpolation on the Permutohedral Lattice for
             Scalable Gaussian Processes

         Sanyam Kapoor * 1 Marc Finzi * 1 Ke Alexander Wang 2 Andrew Gordon Wilson 1

             Abstract
   State-of-the-art methods for scalable Gaussian
   processes use iterative algorithms, requiring fast
   matrix vector multiplies (MVMs) with the co-
   variance kernel. The Structured Kernel Interpo-
   lation (SKI) framework accelerates these MVMs
   by performing effici ...