Struct FunFemConfig 

#[non_exhaustive]
pub struct FunFemConfig {
    pub k: usize,
    pub ncomp: usize,
    pub p_disc: usize,
    pub max_iter: usize,
    pub tol: f64,
    pub seed: u64,
}

Configuration for funFEM discriminative-subspace clustering.

funFEM applies the Fisher-EM algorithm to functional data: it first extracts global FPC scores via fdata_to_pc_1d, then alternates between finding a discriminative subspace (maximising between-class vs within-class scatter) and running a GMM E/M step in that subspace.

Divergence from R funFEM

This is a simplified Fisher-EM implementation. The discriminative subspace is found by computing W_soft^{-1} B_soft via Cholesky inversion followed by an SVD (instead of a proper generalized-eigenvalue solver), because no generalized-eigenvalue crate is used. The multi-pass outer loop re-estimates the subspace at each iteration. This diverges from the iterative schedule in the original paper (Bouveyron & Brunet, 2014) but gives practical cluster recovery on well-separated functional data.

Example

use fdars_core::clustering_advanced::{funfem_cluster, FunFemConfig};
use fdars_core::matrix::FdMatrix;
use std::f64::consts::PI;

let m = 30;
let n = 12;
let t: Vec<f64> = (0..m).map(|i| i as f64 / (m - 1) as f64).collect();
let mut col_major = vec![0.0_f64; n * m];
for i in 0..6 {
    for (j, &tj) in t.iter().enumerate() {
        col_major[i + j * n] = (2.0 * PI * tj).sin();
    }
}
for i in 6..12 {
    for (j, &tj) in t.iter().enumerate() {
        col_major[i + j * n] = (2.0 * PI * tj).sin() + 5.0;
    }
}
let data = FdMatrix::from_column_major(col_major, n, m).unwrap();

let mut cfg = FunFemConfig::default();
cfg.k = 2;
cfg.ncomp = 4;
let result = funfem_cluster(&data, &t, &cfg).unwrap();
assert_eq!(result.cluster.len(), n);

Fields (Non-exhaustive)

Non-exhaustive structs could have additional fields added in future. Therefore, non-exhaustive structs cannot be constructed in external crates using the traditional Struct { .. } syntax; cannot be matched against without a wildcard ..; and struct update syntax will not work.

k: usize

Number of clusters (default: 2). Must be ≥ 1 and ≤ n.

ncomp: usize

Number of global FPC components for the score space (default: 10). Clamped internally to min(n, m).

p_disc: usize

Discriminative subspace dimension (default: 0 = auto = min(k-1, ncomp_eff)). Clamped to ncomp_eff if larger.

max_iter: usize

Maximum outer Fisher-EM iterations (default: 50).

tol: f64

Log-likelihood convergence tolerance (default: 1e-6).

seed: u64

Random seed for k-means++ initialization (default: 42).

Trait Implementations

impl Clone for FunFemConfig

fn clone(&self) -> FunFemConfig

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for FunFemConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for FunFemConfig

fn default() -> Self

Returns the “default value” for a type.

impl PartialEq for FunFemConfig

fn eq(&self, other: &FunFemConfig) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl StructuralPartialEq for FunFemConfig