Tidy tables of a memory-network community result

Usage

# S3 method for class 'net_hon_communities'
hg_get(
  x,
  what = c("states", "physical", "modules", "trials", "first_order", "codelength"),
  ...,
  community = NULL,
  overlapping = FALSE
)

Arguments

x

A net_hon_communities object.

what

Which table:

"states" (default): one row per memory node: node (the memory node, "a -> b"), state (the physical state it belongs to, b), community, flow.
"physical": one row per state x community it appears in: state, community, flow (\(\pi_{i \cap m}\)), share (of that state's flow), n_communities (> 1 marks an overlapping state).
"modules": one row per community: community, n_nodes (memory nodes), n_states, flow, exit_flow, enter_flow.
"trials": one row per search run: run, codelength, n_communities, ari_to_best (adjusted Rand index with the kept partition), best.
"first_order": one row per state: state, its community in the best first-order partition, and its first-order flow.
"codelength": one row per model (memory, first_order): codelength, index_codelength, module_codelength, one_level_codelength, savings_bits, savings_pct, n_communities.

...

Ignored.

community

Integer vector or NULL: keep only these communities (tables "states", "physical", "modules").

overlapping

Logical: for what = "physical", keep only states that appear in more than one community. Default FALSE.

Value

A base data.frame as described under what.

Examples

seqs <- list(c("a", "h", "b", "a", "h", "b", "a"),
             c("c", "h", "d", "c", "h", "d", "c"))
comm <- hg_communities(hon(seqs, max_order = 2L), trials = 2L)
hg_get(comm, what = "modules")
#>   community n_nodes n_states flow exit_flow enter_flow
#> 1         1       5        5    1         0          0
hg_get(comm, what = "physical", overlapping = TRUE)
#> [1] state         community     flow          share         n_communities
#> <0 rows> (or 0-length row.names)