Graphical model for the WSBM. Each weighted edge |$A_{ij}$| (plate) is distributed according to the appropriate edge parameter |$\theta _{z_i,z_j}$| for each observed interaction |$(i,j)$|⁠. In our variational Bayes inference scheme, the WSBM's latent parameters |$z,\theta $| are themselves modeled as random variables distributed according to |$\mu ,\tau $|⁠, respectively. We highlight that the arrow from |$z$| to |$\theta _{z_i,z_j}$| hides the complex relational structure between each |$z_i$|⁠.