Struct NewtonTrustRegion 

pub struct NewtonTrustRegion<ObjFn> { /* private fields */ }

Implementations

impl<ObjFn> NewtonTrustRegion<ObjFn>

where ObjFn: SecondOrderObjective,

pub fn new(x0: Array1<f64>, obj_fn: ObjFn) -> Self

Creates a new Newton trust-region solver.

Arguments

• x0 - The initial guess for the minimum.
• obj_fn - Second-order objective.

pub fn with_tolerance(self, tolerance: Tolerance) -> Self

Sets the convergence tolerance on projected gradient norm (default: 1e-5).

pub fn with_max_iterations(self, max_iterations: MaxIterations) -> Self

Sets the maximum number of iterations (default: 100).

pub fn with_fd_hessian_step(self, fd_hessian_step: f64) -> Self

pub fn with_bounds(self, bounds: Bounds) -> Self

Provides simple box bounds for each coordinate (lower <= x <= upper).

pub fn with_profile(self, profile: Profile) -> Self

pub fn with_hessian_fallback_policy(self, policy: HessianFallbackPolicy) -> Self

Choose what to do when the objective returns SecondOrderSample { hessian: None }: estimate the Hessian by finite-differencing the gradient (default; legacy behavior) or surface a fatal evaluation error. Set to Error when the caller guarantees an analytic Hessian on every call — a single None then signals a routing/contract mismatch instead of silently triggering O(n) extra gradient probes per iteration.

pub fn with_fallback_policy(self, policy: FallbackPolicy) -> Self

Choose whether the trust-region solver may demote to BFGS on step failure. FallbackPolicy::Never keeps the solver inside its analytic-Hessian geometry on rejection; AutoBfgs (the default for Profile::Robust / Aggressive) restarts BFGS from the current best point when no second-order step makes progress.

pub fn with_initial_sample( self, x0: Array1<f64>, sample: SecondOrderSample, ) -> Self

Hand the solver a precomputed (x0, sample) pair so its first evaluation (cost + gradient + Hessian) is served from cache instead of re-running the objective. The sample’s hessian: Option<Array2<f64>> is honored: if None the Hessian is recomputed (or finite-differenced, depending on with_hessian_fallback_policy).

Validation of shape and finiteness is deferred to run(); calling with_initial_sample itself never fails.

pub fn with_initial_trust_radius(self, radius: f64) -> Self

Set the initial trust radius (default: 1.0). Useful when the caller has scale information — e.g. warm-starting from a previous run’s OptimizationDiagnostics.final_trust_radius.

pub fn with_max_trust_radius(self, radius: f64) -> Self

Set the maximum trust radius (default: 1e6).

pub fn with_gradient_tolerance(self, tol: GradientTolerance) -> Self

See Bfgs::with_gradient_tolerance.

pub fn with_observer<O>(self, observer: O) -> Self

where O: OptimizerObserver + 'static,

See Bfgs::with_observer.

pub fn run(&mut self) -> Result<Solution, NewtonTrustRegionError>

Executes the Newton trust-region optimization.

pub fn run_report(&mut self) -> OptimizationReport

Structured report variant of run(). See Bfgs::run_report.

Populates OptimizationDiagnostics.final_trust_radius from the solver’s last observed trust radius so the caller can warm-start a follow-up NewtonTrustRegion::with_initial_trust_radius.