Crate fpcoa

Crate fpcoa 

Source
Expand description

fpcoa: Principal Coordinate Analysis (PCoA) with fixed-rank randomized SVD. Assumptions:

  • dist is dense, strictly symmetric, with zeros on the diagonal. Pipeline (Halko–Martinsson–Tropp range finder + symmetric core):
  1. Build B = -0.5 * J * (D ∘ D) * J (double-centering)
  2. Q = subspace_iteration_full(B, rank = k + oversample, nbiter = q) (annembed/lax)
  3. B_r = Qᵀ * B * Q (r × r, symmetric)
  4. Symmetric eig (nalgebra): B_r = V Λ Vᵀ (Λ sorted ↓)
  5. Zero-out negative eigenvalues (and their eigenvectors)
  6. U ≈ Q * V_k
  7. Coordinates = U * sqrt(Λ_k)

Structs§

FpcoaOptions
Options for randomized PCoA.
PCoAResult
Result of PCoA.

Functions§

pcoa_randomized
Perform PCoA via randomized range finder + symmetric eig on the small core.