Skip to main content

kernel_classify_from_distances

fdars_core::classification::kernel

Function kernel_classify_from_distances 

Source 
pub fn kernel_classify_from_distances(
    func_dists: &[f64],
    y: &[usize],
    scalar_covariates: Option<&FdMatrix>,
    h_func: f64,
    h_scalar: f64,
) -> Result<ClassifResult, FdarError>
Expand description

Kernel classifier from a precomputed functional distance matrix.

Works with any distance matrix (elastic, DTW, Lp, or custom). Bandwidth is selected via LOO-CV if h_func <= 0.

§Arguments

  • func_dists — Flat n × n functional distance matrix (row-major)
  • y — Class labels (length n, 0-indexed)
  • scalar_covariates — Optional scalar covariates (n × p), uses Euclidean distance internally
  • h_func — Functional bandwidth (0 = auto via LOO-CV)
  • h_scalar — Scalar bandwidth (0 = auto)

§Errors

Returns errors if y.len() != n or fewer than 2 classes.