Compute Heat Kernel Signatures (HKS) from a spectral basis

Usage

compute_hks_descriptors(
  basis = NULL,
  values = NULL,
  vectors = NULL,
  q = 16L,
  times = NULL,
  time_mode = c("auto", "fixed"),
  time_range = c(0.1, 50),
  spacing = c("log", "linear"),
  time_scale = 1,
  eigen_tol = 1e-12,
  normalize = c("both", "column", "row", "none"),
  clamp_exponent = -745
)

Arguments

basis: Optional basis list with fields `values` and `vectors`.
values: Optional numeric vector of eigenvalues (overrides `basis$values`).
vectors: Optional matrix (or sparse Matrix) of eigenvectors (overrides `basis$vectors`).
q: Integer number of time points.
times: Optional numeric time grid. If provided, `time_*` arguments are ignored.
time_mode: Either `"auto"` (derive a time range from the spectrum) or `"fixed"` (use `time_range`).
time_range: Length-2 positive numeric vector giving min/max times when `time_mode="fixed"`.
spacing: Either `"log"` (default; typical for HKS) or `"linear"`.
time_scale: Positive scalar multiplier applied to the time grid.
eigen_tol: Positive threshold; eigenpairs with `value <= eigen_tol` are dropped.
normalize: Descriptor normalization: `"both"` (default), `"column"`, `"row"`, or `"none"`.
clamp_exponent: Lower clamp for the exponent (default `-745`), used to avoid denormal underflow in `exp(-λ t)`.

Value

A numeric matrix of dimension N x q.