§6.1, pp. 28–29 — the starting point xⁱ⁽⁰⁾ = sⁱ⁽⁰⁾ = Tⁱe₁ⁱ, τ⁽⁰⁾ = κ⁽⁰⁾ = 1 lies in N(1)
ProvedConicQuadIPM.Complementarity.starting_pointcentral-pathconic-optimizationinterior-pointp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1second-order-cone
Let be as in §3 and define the starting point
where is the first unit vector of . Then
The algorithm of the paper starts from this point, so it starts on the central path.
Formalization Note. The page also sets , which plays no role in and is omitted. Block dimensions (WellFormed) are a disclosed hypothesis.
Preamble
import Mathlib import Definitions.Def_ConicQuadIPM_Complementarity_Setting open Matrix
Formal statement
namespace ConicQuadIPM.Complementarity
theorem starting_point {k : ℕ} (kind : Fin k → ConeKind) (n : Fin k → ℕ)
(hwf : WellFormed kind n) :
Nbhd kind n 1
(fun i => Tmat (kind i) (n i) *ᵥ (e1 : Fin (n i) → ℝ)) 1
(fun i => Tmat (kind i) (n i) *ᵥ (e1 : Fin (n i) → ℝ)) 1 := by sorry
end ConicQuadIPM.Complementarity
Source
Andersen, Roos & Terlaky, On implementing a primal-dual interior-point method for conic quadratic optimization, Math. Program. (2003); authors' preprint of 18 Dec 2000, §6.1, pp. 28–29
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.