Skip to contents

EXPERIMENTAL partitioned Walktrap on a file-backed fuzzy graph

Source: R/massive_walktrap.R
massive_walktrap.Rd

Runs the existing exact Walktrap kernel on bounded induced graph blocks, contracts their communities, and repeats on the contracted graph. This is an approximation when more than one block is needed: cross-block walks are absent from the first level, and results depend on vertex order. A graph that cannot contract fails explicitly, leaving partial labels for inspection. Only CPU is supported. Exact in-memory graph_cluster() is unchanged.

Usage

massive_walktrap(
  graph,
  output,
  steps = 4L,
  chunk_rows = 4000L,
  memory_limit = "1GB",
  backend = "cpu"
)

Arguments

graph

A symmetric file-backed fuzzy graph.

output

New .clusters.u32 path.

steps

Random-walk length.

chunk_rows

Maximum vertices per induced graph block.

memory_limit

Host working-memory budget.

backend

Must be "cpu".

Value

File-backed Walktrap labels and approximation metadata.