Deep kernel processes

               Laurence Aitchison 1 Adam X. Yang 1 Sebastian W. Ober 2

            Abstract                already exist. In these methods, which are closely re-
                                lated to infinite Bayesian neural networks (Lee et al., 2017;
   We define deep kernel processes in which positive     Matthews et al., 2018; Garriga-Alonso et al., 2018; Novak
   definite Gram matrices are progressively trans-      et al., 2018), we take an initial kernel (usually ...