Global inducing point variational posteriors for Bayesian neural networks and
             deep Gaussian processes

                     Sebastian W. Ober 1 Laurence Aitchison 2

             Abstract                 specify a family of approximate posteriors over the weights,
                                  with the classical approach being independent Gaussian dis-
    We consider the optimal approximate posterior        tributions over each individual weight (Hinton & Van Camp,
    over the ...