All the reviewers agreee that the paper proposed an interesting idea wrt structured variational approximations in DGPs but raised concerns with regards to the experimental evaluation. Most of these were addressed in the rebuttal and, in agreement with the reviewers, I recommend acceptance.