Alphabeta Math
ExampleConstruction: Literature-sourcedVerification: AI-adaptedPipeline-generatedprecheck pass
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

A full SQP step is rejected by the L1 merit test and repaired by SOC

Statement

Let h(x)=x12+x22−1=0 and f(x)=2(x12+x22−1)−x1. At a feasible point x=(cos⁡θ,sin⁡θ)=(c,s), use multiplier estimate ν=−3/2 and SQP Hessian model B=I. For nonzero sufficiently small θ, the SQP step p=(s2,−sc) has predicted decrease Dρ=s2>0, but its full step increases the L1 merit for every ρ>0. The minimum-norm SOC correction makes the full corrected step pass the Armijo test with σ=1/4 for every fixed ρ>0 and all sufficiently small nonzero θ.

Facts & Assumptions

Given: The equality-constrained SQP and L1 merit definitions, the SOC formula, and c=cos⁡θ, s=sin⁡θ.

[F1]

The linearized quadratic subproblem in sequential quadratic programming imposes h(x)+Dh(x)p=0 in the subproblem.

[F2]

L1 exact-penalty merit and predicted decrease for SQP defines the merit and predicted decrease Dρ.

[F3]

Minimum-norm second-order correction for identified active constraints sets w=−Dh(x)T(Dh(x)Dh(x)T)−1h(x+p).

[F4]

Second-order correction reduces the active constraint residual to cubic order gives the local quadratic correction size and cubic active-residual order under its smooth full-rank hypotheses.

[F5]

L1 merit globalization with finite penalty and backtracking budgets tests the full SOC candidate against the Armijo threshold.

[A1]

θ is sufficiently small and nonzero, ρ>0 is fixed, B=I, the multiplier estimate is −3/2, and σ=1/4.

Proof

1.1givenF1A1algebra

At x=(c,s), h(x)=0, Dh(x)=2(c,s), and ∇f(x)=4(c,s)−(1,0). The unit tangent t=(−s,c) satisfies tT∇f(x)=s. The QP constraint gives p∈span⁡(t); minimizing ∇f(x)Tp+12∥p∥22 on this line gives p=−st=(s2,−sc). Its QP multiplier is −(4−c)/2, since p+∇f(x)=(4−c)(c,s).

1.2A1F3algebra

Here Dh(x)Dh(x)T=4, so [F3] gives w=(−cs2/2,−s3/2) and ∥w∥2=s2/2. Direct substitution into the quadratic constraint yields h(x+p+w)=s44. The corrected first coordinate is c+s2(1−c/2), so the corrected objective change is −s2(1−c/2)+s4/2. Adding the penalty contribution gives ψρ(x+p+w)−ψρ(x)=−s2(1−c/2)+s4(1/2+ρ/4).

2.1A1F2step 1.1algebra

Since xTp=0 and ∥p∥22=s2, h(x+p)=s2. Also ∇f(x)Tp=−s2, so [F2] gives Dρ=s2. The objective changes by s2, while the equality violation changes from zero to s2; hence the full-step merit change is (1+ρ)s2>0, which fails the Armijo right side −σs2<0.

3.1A1F4F5step 1.2algebra∎

Divide the corrected merit change plus σDρ by s2. With σ=1/4, the result is −1+c/2+s2(1/2+ρ/4)+1/4, which tends to −1/4 as θ→0. It is therefore negative for all sufficiently small nonzero θ, proving the SOC Armijo test passes. The explicit O(s2) correction norm and O(s4) residual satisfy the quadratic-size and cubic-residual bounds of [F4].

Depends on

Used by

Nothing in the library uses this result yet.

Dependency tree · two levels

10 results within two dependency steps of this one, each drawn at its shortest distance from it. An arrow runs from a result to what uses it, so the chart reads left to right and ends at this result, which carries a heavier outline. Every node is a link to that result. Click elsewhere on the chart to enlarge it.

Sources