Stationarity matrix
DefinitionMatrixLeeSidford_Supplement_statMatinterior-point-methodsp2o-gran-per-chapterp2o-plan-bookp2o-v1self-concordant-barrierssemidefinite-programming
Let be feasible, a weight with Cholesky factor , and let and its trace adjoint be the maps of Definition 4.3 at . The stationarity matrix at the query state is the matrix
At the optimal weight stationarity gives ; the algorithm's G-steps drive an estimate of toward .
Formalization Note is the Cholesky factor of , as printed; is a natural-number exponent ( in Part II).
Definition code
import Mathlib
import Definitions.Def_MatrixLeeSidford_Lmap
import Definitions.Def_MatrixLeeSidford_Ladj
import Definitions.Def_MatrixLeeSidford_slack
import Definitions.Def_MatrixLeeSidford_Supplement_cholFactor
open Matrix
namespace MatrixLeeSidford.Supplement
/-- The **stationarity matrix** `R = R_G(y) := Lᵀ L_y^*(L_y(G)^{s−1}) L` at a query state `(y, G)`, where
`L` is the Cholesky factor of `G` (Song–Zhang, *A Matrix Lee–Sidford Barrier and Faster Semidefinite Programming*, STOC 2027 submission manuscript of October 31, 2026 (sha256 7c638b45…412c1), Definition 8.1, p.53).
**Formalization Note** `L_y = Lmap A (S(y))` and `L_y^* = Ladj A (S(y))` are the published maps at the
slack `S(y)`; `L` is `cholFactor G`; the power `s − 1` is a natural-number matrix power (Part II has
`s ≥ 2`). -/
noncomputable def statMat {n m : ℕ} (A : Fin m → Matrix (Fin n) (Fin n) ℝ)
(C : Matrix (Fin n) (Fin n) ℝ) (s : ℕ) (y : EuclideanSpace ℝ (Fin m))
(G : Matrix (Fin m) (Fin m) ℝ) : Matrix (Fin m) (Fin m) ℝ :=
(cholFactor G)ᵀ * Ladj A (slack A C y) ((Lmap A (slack A C y) G) ^ (s - 1)) * cholFactor G
end MatrixLeeSidford.Supplement
Source
Song–Zhang, A Matrix Lee–Sidford Barrier and Faster Semidefinite Programming, STOC 2027 submission (manuscript of October 31, 2026, sha256 7c638b45…412c1), Definition 8.1, p.53