Hello everybody, In my research I am working on networks community structure and I am interested in investigating the relationship between community structure and missing links.
Thus I applied hrg.consensus and hrg.predict to subgraph representing network communities (found using fastgreedy.community). In order to extract the subgraph I used induced.subgraph (initially without specifing an implementation). Here it is the instructions I used: sub_graph <- induced.subgraph(graph=g_n1, v=unlist(fg_communities[k])) g_n1_hrg_con <- hrg.consensus(sub_graph) g_n1_miss_link <- hrg.predict(graph = sub_graph, hrg = g_n1_hrg_con) As reported in the help in this case induced.subgraph choose heuristically what implementation to use. As a result I obtain that the $edge matrix, element of the list returned by hrg.predict, stores vertices that don't belong to the subgraph. I am working with several networks of different size and from my tests I found that it doesn't happen every time. My supposition is that in sometimes induced.subgraph use "copy_and_delete" and some others "create_from_scratch". WHEN it choose "copy_and_delete" it store somewhere the information about the vertices of the main graph, SO when I use subgraph as argument for hrg.consensus and hrg.predict it considers the "external" vertices too (those not belonging to the subgraph). I have some evidences that using "create_from_scratch" this issue doesn't occur. I would like to know if this assumption is correct and if I can avoid vertices not belonging to a subgraph to show in $edges matrix using "create_from_scratch" as a imp value in induced.subgraph. Thank you very much in advance. Matteo Orrù I think to have found a strange behaviour of hrg.consensus and hrg.predict while using the _______________________________________________ igraph-help mailing list [email protected] https://lists.nongnu.org/mailman/listinfo/igraph-help
