Crate spectrafit_newton_cg 

spectrafit-newton-cg — matrix-free Newton-CG (Steihaug–Toint) trust region.

One method on the graph-agnostic spectrafit-trust-region framework: it supplies the SteihaugStep subproblem solver — truncated conjugate gradients that never form JᵀJ, using only matrix-free H·v = J̃ᵀ(J̃·v) products — and a thin minimize entry point; the framework owns the Δ-radius control loop.

This is the large-scale lever: the per-iteration cost scales with the number of residuals, not p², so it stays cheap when the parameter count is large or JᵀJ is ill-conditioned (the regime where forming the normal equations squares the condition number).

Structs

Report

Outcome of a solve. The optimised parameters live in the problem (read via TrustRegionProblem::params); this only carries diagnostics.

SteihaugStep

Matrix-free Newton-CG (Steihaug–Toint) trust-region subproblem solver (stateless).

TrustRegionConfig

Tuning for a single Δ-radius trust-region solve.

Enums

Termination

Why the solve stopped.

Traits

TrustRegionProblem

A weighted nonlinear least-squares problem driven by the trust-region core.

Functions

minimize

Minimise ½‖r(p)‖² over the free parameters of problem with the matrix-free Newton-CG (Steihaug–Toint) trust-region method. On return the problem holds the best parameters.