Feature requests for LayeredBlockState: different rec_types for different layers + PPBlockState as base state
Hello,
a) as it is currently not possible to pass different rec_types onto LayeredBlockState, adding this feature would potentially allow for better fit for models with layers with different types of edge weight distributions. Based on a thread on the graph-tool mailing list, this should become available in a future release under "StackedBlockState".
b) just using this opportunity to ask for another feature: while the basic PPBlockState already exists, ideally - as there are probably many who'd be interested in inferring assortative communities - one would be able to pass the planted partition model as a base state for LayeredBlockState - that is, use PPBlockState with layers, edge weights and/or by fitting an overlapping model.
Many thanks, Arttu