Detect probabilistic extrema in an rdgraph fitted surface

Retained in gflow because this is an extrema computation API, even though its input is the archived riem.dcx fit class now produced by gflowx. The function filters candidate vertices by extremum probability and effective degree, then measures the hop radius over which the probabilistic extremum condition persists.

Usage

compute.pextrema.nbhds(
  dcx,
  extremp_quantile = 0.95,
  hop_extremp_threshold = 0.9,
  eff_degree_quantile = 0.1,
  max_hop = 20L,
  compute_hop_idx = FALSE
)

Arguments

dcx

A riem.dcx object, normally returned by gflowx::fit.rdgraph.regression().

extremp_quantile

Quantile in [0, 1] used to select initial maximum and minimum candidates.

hop_extremp_threshold

Probability threshold in (0, 1] used when expanding hop neighborhoods.

eff_degree_quantile

Quantile in [0, 1] below which weakly connected candidate vertices are discarded.

max_hop

Positive integer maximum hop distance to explore.

compute_hop_idx

Whether to add the classical hop index computed by compute.extrema.hop.nbhds().

Value

A data frame with one row per selected extremum and columns vertex, value, rel_value, type, extremp, eff_degree, hop_extremp_radius, hop_idx, and label.

See also

maxp(), minp(), compute.extrema.hop.nbhds()