Abstract

Recent research has shown the potential utility of probability distributions designed through hierarchical constructions which are conditionally Gaussian. This body of work is placed in a common framework and, through recursion, several classes of deep Gaussian processes are defined. The resulting samples have a Markovian structure with respect to the depth parameter and the effective depth of the process is interpreted in terms of the ergodicity, or non-ergodicity, of the resulting Markov chain.

Citation information

Dunlop, Matthew & Girolami, Mark & M. Stuart, Andrew & Teckentrup, Aretha. (2017). How Deep Are Deep Gaussian Processes?

Turing affiliated authors