Enum InitialMetric 

pub enum InitialMetric {
    Identity,
    Scalar(f64),
    Diagonal(Array1<f64>),
    DenseInverseHessian(Array2<f64>),
}

How to seed the BFGS inverse-Hessian approximation H_0^{-1}. Replaces (and supersedes) the previous with_initial_inverse_hessian thinking by giving callers a clean choice between scaled-identity resets and full dense seeds.

Default behavior (when no with_initial_metric is set) is Identity: BFGS uses its internal scaled-identity initialization.

Variants

Identity

Default: scaled identity (BFGS picks the scale internally).

Scalar(f64)

H_0^{-1} = scale * I. Useful when the caller has a single magnitude estimate (e.g. from a previous run’s gradient norm or from a known curvature scale) but no per-coordinate information.

Diagonal(Array1<f64>)

H_0^{-1} = diag(diag). Per-coordinate scaling — the typical shape for penalized likelihoods where the penalty matrix is near-diagonal in the smoothing-parameter coordinate system.

DenseInverseHessian(Array2<f64>)

A complete dense H_0^{-1}. The matrix is validated for shape and finiteness at run() time; symmetry is not enforced (BFGS’s update preserves whatever symmetry the seed has).

Trait Implementations

impl Clone for InitialMetric

fn clone(&self) -> InitialMetric

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for InitialMetric

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

Formats the value using the given formatter.