DNN weights (choose method), distance-based ranking

Use method = "exact" for the exact finite-population rank-weight rule and method = "asymp" for the asymptotic full-rank approximation. Fixed top-k kernels are available in both cases. The package treats them as implementation-supported generalizations; the accompanying manuscript's current derivations are written for k = 1.

Usage

weights_dnn(
  X,
  X0 = NULL,
  x0 = NULL,
  distance = "l2",
  k,
  s,
  method = c("exact", "asymp"),
  normalize = TRUE,
  ctx = NULL,
  return = c("W")
)

Arguments

X: numeric matrix (n x p)
X0: numeric matrix (m x p); mutually exclusive with x0
x0: numeric length-p single target (convenience for m=1)
distance: distance spec or registered name (e.g., "l2")
k: integer; neighbors used in each s-subset (1 <= k <= s)
s: integer; U-statistic order / subset size
method: one of "exact" or "asymp"
normalize: logical; renormalize columns to sum 1
ctx: optional context
return: character vector of extras to include; any of c("W","ord","ranks","dx")

Value

A weights_result object.

Examples

set.seed(1)
x_mat <- matrix(rnorm(20), 10, 2)
x0 <- c(0, 0)
out <- weights_dnn(x_mat, x0 = x0, k = 1L, s = 3L, method = "exact")
dim(out$W)
#> [1] 10  1