Construct a kNN anchor graph and Laplacian

Construct a kNN anchor graph and Laplacian

Usage

dkge_anchor_graph(
  anchors,
  k = 10,
  sigma = NULL,
  weight_mode = c("heat", "normalized", "binary"),
  type = c("mutual", "normal", "asym")
)

Arguments

anchors

\(Q \times 3\) matrix of anchor coordinates.

k

Number of neighbours used in the graph (default 10).

sigma

Optional Gaussian length-scale (mm). If `NULL`, the `adjoin` defaults are used.

weight_mode

Edge weighting scheme passed to [adjoin::graph_weights()]. Defaults to `"heat"`.

type

Graph symmetrisation strategy (see [adjoin::graph_weights()]); `"mutual"` helps enforce symmetry.

Value

A list containing the neighbour graph, sparse adjacency `W`, degree matrix `D`, and Laplacian `L`.

Examples

anchors <- matrix(rnorm(40 * 3), 40, 3)
g <- dkge_anchor_graph(anchors, k = 3)
names(g)
#> [1] "graph" "W"     "D"     "L"