Leiden or Louvain community detection on an igraph — leiden

Modularity-based Leiden (Traag et al.) or Louvain (Blondel et al.) community detection, run in Rust. Mirrors igraph::cluster_leiden(objective_function = "modularity"), but the result depends on seed only (not on the global RNG or the number of threads). Communities are numbered by size, 1 being the largest.

Usage

leiden_clustering(
  graph,
  resolution = 1,
  algorithm = c("leiden", "louvain"),
  beta = 0.01,
  n_iterations = NULL,
  min_gain = 0,
  local_moving = c("queue", "coloured"),
  seed = 42L
)

Arguments

graph: igraph. Undirected. If it has an edge attribute weight, it is used, otherwise all edges get weight 1.
resolution: Numeric. Modularity resolution parameter. Higher values give more communities.
algorithm: String. One of c("leiden", "louvain").
beta: Numeric. Randomness of the Leiden refinement merges. Ignored for Louvain.
n_iterations: Optional integer. Number of full iterations, each starting from the previous partition. A negative value iterates until the partition is stable. If NULL, 5L for Leiden and 2L for Louvain, which gave the best modularity across synthetic graphs.
min_gain: Numeric. With a negative n_iterations, also stop once an iteration gains no more than this in modularity.
local_moving: String. One of c("queue", "coloured"). "queue" is Traag's sequential queue. "coloured" moves the nodes of one colour class in parallel, which only kicks in on larger levels of the graph.
seed: Integer. Random seed.

Value

Integer vector with the community of each node (1-based, 1 is the largest community).

References

Traag, Waltman & van Eck, Sci Rep, 2019; Blondel, et al., J Stat Mech, 2008

Examples

g <- igraph::make_graph("Zachary")
table(leiden_clustering(g))
#> 
#>  1  2  3  4 
#> 12 11  6  5