“…Automating the model design cycle [ 2 ] under the Bayesian formalism has been the goal of many probabilistic programming languages [ 8 , 9 , 10 , 11 , 12 , 13 , 14 ]. This paper focuses on message passing-based approaches, which leverage the conditional independencies in the model structure for performing probabilistic inference, e.g., [ 27 , 28 , 29 , 30 ], which will be formally introduced in Section 3.2 .…”