Peripheral seed index for Gonzalez farthest-first selection

Returns the index of a single deterministic peripheral seed, the starting point used by Gonzalez() under the corresponding seed strategy. Useful when composing a custom selection pass.

Usage

MaxMinSeed(
  d = NULL,
  points = NULL,
  method = c("peripheral", "centroid", "random_furthest", "diameter", "anti_medoid",
    "medoid", "rowsum", "rownorm")
)

Arguments

d: A dist object or square symmetric numeric matrix. Ignored when points is supplied.
points: Optional N x dim numeric coordinate matrix; when supplied the seed is computed from coordinates in O(N) memory. Required for the "centroid" anchor, which has no distance-matrix form.
method: Anchor name (see above). Default "peripheral".

Value

Integer seed index in [1, N].

Details

Anchors:

"centroid": The point farthest from the coordinate mean (argmax ||x - x_bar||), an O(N * dim) approximate diameter endpoint. Computed from coordinates, so it requires points; it is unavailable on the distance-matrix path, where "peripheral" serves the same role.
"peripheral": Two sweeps: the point furthest from point 1, then the point furthest from that (a diameter-endpoint approximation), in O(N). The only anchor reachable from a distance-column oracle (the function path of Gonzalez()).
"random_furthest": The point furthest from a random pivot, in O(N). The pivot is drawn from a fixed internal seed, isolated from the ambient RNG, so the index is reproducible across sessions and machines.
"diameter": A row endpoint of the diameter pair (the maximum pairwise distance). Degenerate data (d_max <= 0) falls back to 1.
"anti_medoid": The point furthest from the 1-median (medoid).
"medoid": The 1-median itself, which.min(rowSums(d)).
"rowsum": The point maximising the sum of distances to all others (the 1-anti-median).
"rownorm": The point maximising sqrt(sum d^2), the L2 counterpart of "rowsum".

Examples