• Changes to mlvsbm_estimate_generalized_network():
• Fixed bug when setting the number of clusters.
• Added init_points and Qmax to the fit_options argument.
• The default algorithm is now slower but handles better underfitting (fit_option\$init_points == “all”). Users may fix it to any other value to revert to the old behavior.
• When fixing the number of clusters, the algorithm will better explore neighbors models.
• Returned blocks are reordered so that the affiliation parameters (gamma) is as “diagonal” as possible.
• Added support for generalized multilevel networks:
• This includes three new user functions: mlvsbm_simulate_generalized_network(), mlvsbm_create_generalized_network() and mlvsbm_estimate_generalized_network().
• This function added the following features: “poisson” distribution of the interactions to deal with count data, support for network of more than two levels (temporal networks e.g.), support for networks with nodes with no affiliation.
• Minor bug fix
• Changed plot function and method of FitMLVSBM object to graphon view type.
• Improved initialization for networks with missing data.
• Fixed a bug for directed networks.
• Added S3 generic function print, plot, predict and coef for FitMLVSBM object.
• R6 class methods are better documented.
• Reverse the sbm dependency to blockmodels for future sbm compatibility.
• Fixed a bug with NA interactions on user functions.
• added sbm package dependency for better initialization.
• Added options for mlvl_estimate_network() function.
• It is now possible to simulate networks with no isolated nodes.
• Fixed bug with parallel for Windows.
• Finished user functions.
• Tutorial.
• Added a NEWS.md file to track changes to the package.