Network analysis has a crucial need for tools to compare networks and assess the significance of differences between networks. We propose a principled statistical approach to network comparison that approximates networks as probability distributions on negatively curved manifolds. We outline the theory, as well as implement the approach on simulated networks.