Struct QuadRowScale 

pub struct QuadRowScale {
    pub index: usize,
    pub name: String,
    pub curvature: Number,
    pub linear: Number,
    pub rhs: Number,
    pub jac_at_x0: Number,
    pub scale: Number,
    pub mismatch: Number,
}

The coefficient magnitudes of one quadratic constraint row, read off the .nl without reference to any point, paired with what the gradient sample at x0 made of it.

curvature is ‖Q‖_∞ (the largest absolute row sum of the row’s Hessian), which is Gershgorin’s bound on λ_max(Q) and is the exact quantity §8’s second-stage row scale eᵢ = 1/max(‖Qᵢ‖_∞, ‖aᵢ‖_∞, |bᵢ|) is built from. It is an upper bound on the curvature, not the curvature.

Fields

index: usize

name: String

curvature: Number

‖Q‖_∞ — see the type docs.

linear: Number

‖a‖_∞ over the .nl linear section plus the degree-1 terms the writer folded into the nonlinear tree.

rhs: Number

|b| — the finite bound the row is written against, shifted by the folded constant. A range row reports the larger magnitude.

jac_at_x0: Number

‖∇g(x0)‖_∞ — what gradient-based scaling actually samples.

scale: Number

The factor that sample assigns the row.

mismatch: Number

rhs / curvature. §8’s statistic: the qcqp1500-1c right-hand sides are 1.58e5–1.80e5 against λ_max(Qᵢ) ≈ 1.6e3, a 100× mismatch that biases sᵢ = −gᵢ(x) and hence the −sᵢ/λᵢ KKT diagonal. Zero when the curvature is zero.

Trait Implementations

impl Clone for QuadRowScale

fn clone(&self) -> QuadRowScale

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for QuadRowScale

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

Formats the value using the given formatter.