In this paper we introduce deep gaussian process (GP) models. Deep GPs are a deep belief network based on Gaussian process mappings. The data is modeled as the output of a multivariate gp. The inputs to that Gaus