On 25.08.2017 16:29, Valery Topinsky wrote:

Hello.  As I understand, edge multiplicities is kind of number of edges between two nodes.  And we can define it via 'eweight' attribute in BlockState. In some way (may be I'm wrong)  if I differentiate initial edges by assigning to them different 'weights' via 'eweight', I  will have also weighted graph model.  What is the difference between edge covariates and edge multiplicities  besides the generative model priors? Or I can consider them as a special simple case of more general 'edge covariates' of 'recs' attribute?  Is there any discussion about this? 

If you specify the "eweight" parameter, you define a multigraph with the given edge multiplicities. It is equivalent to putting parallel edges on the graph. The "recs" attributes go beyond this, and they add to all edges (even the multiple ones) an additional edge covariate, which can be discrete, but also continuous. Networks with discrete edge covariates can be modeled either as a multigraph (via eweight) or a simple graph with edge covariates. The only difference is the functional form the model has, which is not identical. Best, Tiago -- Tiago de Paula Peixoto <tiago@skewed.de>