Probabilistic Graphical Models in JAX. This is a work in progress. Currently we only support belief propagation in (loopy) Gaussian PGMs. In the case of chain structured graphs with linear-Gaussian potentials, this gives the same result as Kalman smoothing.