Module community

Expand description

Community-detection algorithms (ALGO-CO-*). Phase 1: modularity (Newman-Girvan modularity of a partition). Phase 4: louvain multilevel community detection, leiden (Traag-Waltman-van Eck 2019), label_propagation (Raghavan-Albert-Kumara 2007 + Traag-Šubelj 2023 fast variant), fluid_communities (Parés et al. 2017), edge_betweenness_community (Girvan-Newman 2002), fast_greedy_modularity (Clauset-Newman-Moore 2004), walktrap (Pons-Latapy 2005), leading_eigenvector (Newman 2006), infomap (Rosvall-Bergstrom 2008), spinglass (Reichardt-Bornholdt 2006). Helpers: community_to_membership (cut a dendrogram at k merges), reindex_membership (densify a membership vector to 0..k-1 by first occurrence).

Structs§

CommunityToMembershipResult: Result of community_to_membership: per-vertex cluster ids densified to 0..k, plus the size of each cluster in the same order.
CommunityVoronoiResult: Result of community_voronoi.
EdgeBetweennessResult: Result of edge_betweenness_community.
FastGreedyResult: Result of fast_greedy_modularity / fast_greedy_modularity_weighted.
FluidOptions: Full option bag for fluid_communities_with_options.
FluidResult: Result of fluid_communities / fluid_communities_with_options.
InfomapResult: Result of an Infomap run.
LeadingEigenvectorResult: Result of the leading eigenvector community detection.
LeidenOptions: Full set of options for leiden_with_options. Construct via LeidenOptions::default() and then mutate the fields you care about, mirroring the way upstream C exposes per-parameter defaults.
LeidenResult: Result of a Leiden run.
LouvainResult: Result of a Louvain run.
LpaOptions: Full option bag for label_propagation_with_options.
LpaResult: Result of a label-propagation run.
ReindexMembershipResult: Result of reindex_membership: a densified membership vector [0, k), the old cluster id behind each new id (so new_to_old[i] is the original id that now maps to i), and the number of distinct clusters k = new_to_old.len().
SpinglassOptions: Options for spinglass community detection.
SpinglassResult: Result of spinglass community detection.
SplitJoinDistance: Asymmetric split-join distances between two community partitions.
WalktrapOptions: Tuning knobs for walktrap_with_options.
WalktrapResult: Result of walktrap / walktrap_weighted / walktrap_with_options.

Enums§

CommunityComparison: Which partition-comparison measure compare_communities returns.
LeidenObjective: Quality function (objective) that Leiden optimises. Mirrors igraph_leiden_objective_t.
LpaVariant: Which variant of label propagation to run. Mirrors igraph_lpa_variant_t.
SpinglassUpdateRule: Update rule for the spinglass null model.

Constants§

FLUID_DEFAULT_MAX_ITERATIONS: Default safety cap on outer iterations for fluid_communities / fluid_communities_with_options.
LEIDEN_DEFAULT_BETA: Default randomness used in the refinement step — same default as upstream igraph_community_leiden_simple (β = 0.01).
LEIDEN_DEFAULT_ITERATIONS: Default number of outer iterations. The reference paper and upstream both note that two iterations are usually enough.
SPINGLASS_DEFAULT_SPINS: Default number of spins for spinglass.
WALKTRAP_DEFAULT_STEPS: Default random-walk length matching the igraph reference (steps = 4).

Functions§

community_to_membership: Cut a binary dendrogram at steps merges, returning the resulting membership and cluster-size vectors.
community_voronoi: Voronoi-based community detection.
compare_communities: Compare two community membership vectors over the same vertex set.
edge_betweenness_community: Run edge-betweenness community detection on graph.
edge_betweenness_community_weighted: Run weighted edge-betweenness community detection on graph with per-edge weights.
fast_greedy_modularity: Run fast greedy modularity community detection on graph (unweighted).
fast_greedy_modularity_weighted: Run fast greedy modularity community detection on graph with edge weights.
fluid_communities: Run Fluid Communities with default options (seed = 0, max_iterations = FLUID_DEFAULT_MAX_ITERATIONS).
fluid_communities_with_options: Run Fluid Communities with full option control.
infomap: Run Infomap with the default options (unweighted, 1 trial).
infomap_weighted: Run Infomap with per-edge weights (1 trial).
infomap_with_options: Run Infomap with full control over parameters.
label_propagation: Run label propagation with the default options (LpaVariant::Fast, no initial labels, no fixed vertices, seed 0).
label_propagation_weighted: Run label propagation with per-edge weights (default variant, no initial labels, no fixed vertices, seed 0).
label_propagation_with_options: Run label propagation with the full option bag.
le_community_to_membership: Cut an incomplete dendrogram after steps merges, starting from an initial (non-singleton) cluster assignment.
leading_eigenvector: Detect communities using Newman’s leading eigenvector method.
leading_eigenvector_weighted: Weighted leading eigenvector community detection (convenience wrapper).
leiden: Run Leiden with the default modularity objective, γ = 1, β = 0.01, two iterations, seed 0.
leiden_weighted: Run Leiden with per-edge weights (modularity objective, γ = 1, β = 0.01, two iterations, seed 0).
leiden_with_options: Run Leiden with the full set of options.
louvain: Run Louvain with the default options (γ = 1, unweighted, deterministic seed 0).
louvain_weighted: Run Louvain with per-edge weights (γ = 1, deterministic seed 0).
louvain_with_options: Run Louvain with the full set of options.
modularity: Modularity of graph with respect to community assignment membership.
modularity_directed: Directed modularity (Leicht-Newman, ALGO-CO-001b).
modularity_matrix: Compute the modularity matrix B of a graph.
modularity_weighted: Weighted modularity (ALGO-CO-001c).
reindex_membership: Relabel a membership vector so cluster ids run contiguously over 0..k, where k is the number of distinct clusters. The partition is unchanged: two vertices share a cluster in the output iff they shared one in the input. New ids are assigned in first-occurrence order — the first cluster id you encounter when sweeping left to right becomes 0, the next new one becomes 1, and so on.
spinglass: Spinglass community detection with default options.
spinglass_weighted: Spinglass community detection with edge weights.
spinglass_with_options: Spinglass community detection with full options.
split_join_distance: Compute the two asymmetric projection distances between comm1 and comm2.
walktrap: Run Walktrap on an unweighted, undirected graph with default steps = 4.
walktrap_weighted: Run Walktrap on a weighted, undirected graph with default steps = 4.
walktrap_with_options: Run Walktrap with a custom WalktrapOptions.