Target expectation and the Gram and objective matrices
DefinitionRelDecoding_Supplement_targetExpectationoperator-algebrasp2o-gran-per-chapterp2o-plan-paperp2o-v1quantum-informationrelative-decoding
Let be a unital -homomorphism, the normalized traces, and with .
- For a self-adjoint target observable the target expectation is
(eq. (114)). 2. For a guide family with , the target Gram and objective matrices are and (eq. (115)). 3. The source Gram and objective matrices are and (App. C.1).
is the energy (or score) of the target guide state; by Proposition 5.1 its optimum over a guide space is a generalized eigenvalue problem for .
Formalization Note At the quotient is (Lean's ); the source uses it only for .
Definition code
import Mathlib
import Definitions.Def_LIDT_MIPStarRE_Quantum_FiniteMatrix_NormalizedTrace
import Definitions.Def_RelDecoding_markedSpace
set_option autoImplicit false
namespace RelDecoding.Supplement
open LIDT.MIPStarRE.Quantum RelDecoding
variable {nA nB : Type*} [Fintype nA] [DecidableEq nA] [Fintype nB] [DecidableEq nB] {s : ℕ}
/-- The **target expectation** `E_B(O_B; X_A) := τ_B(X_B∗ O_B X_B) / τ_B(X_B∗ X_B)` with
`X_B := π(X_A)`. Shang, Stilck França, *From Simple Sources to Quantum Advantage: Homomorphic
Polynomial Transduction via Relative Decoding*, arXiv:2609.35101 (held version, sha256 d2be7859…),
eq. (114), p.28.
**Formalization Note.** `τ_B` is `normalizedTrace`. The source uses it only for `X_B ≠ 0`; at
`X_B = 0` Lean's `x / 0 = 0` gives `0`. -/
noncomputable def targetExpectation {a : Fin s → Matrix nA nA ℂ}
(π : genAlg a →⋆ₐ[ℂ] Matrix nB nB ℂ) (OB : Matrix nB nB ℂ) (XA : genAlg a) : ℂ :=
normalizedTrace (star (π XA) * OB * π XA) / normalizedTrace (star (π XA) * π XA)
/-- The **target Gram matrix** `(G_π)_{ij} := τ_B(Y_i∗ Y_j)`, `Y_j = π(F_j)`, of a guide family
`F₀, …, F_{R−1}`. Eq. (115), p.28. -/
noncomputable def targetGram {a : Fin s → Matrix nA nA ℂ} (π : genAlg a →⋆ₐ[ℂ] Matrix nB nB ℂ)
{R : ℕ} (F : Fin R → genAlg a) : Matrix (Fin R) (Fin R) ℂ :=
Matrix.of fun i j => normalizedTrace (star (π (F i)) * π (F j))
/-- The **target objective matrix** `(K_π)_{ij} := τ_B(Y_i∗ O_B Y_j)`, `Y_j = π(F_j)`. Eq. (115),
p.28. -/
noncomputable def targetObjective {a : Fin s → Matrix nA nA ℂ}
(π : genAlg a →⋆ₐ[ℂ] Matrix nB nB ℂ) (OB : Matrix nB nB ℂ) {R : ℕ} (F : Fin R → genAlg a) :
Matrix (Fin R) (Fin R) ℂ :=
Matrix.of fun i j => normalizedTrace (star (π (F i)) * OB * π (F j))
/-- The **source Gram matrix** `(G_A)_{ij} := τ_A(F_i∗ F_j)`. App. C.1, pp.72–73. -/
noncomputable def sourceGram {a : Fin s → Matrix nA nA ℂ} {R : ℕ} (F : Fin R → genAlg a) :
Matrix (Fin R) (Fin R) ℂ :=
Matrix.of fun i j => normalizedTrace (star (F i : Matrix nA nA ℂ) * (F j : Matrix nA nA ℂ))
/-- The **source objective matrix** `(K_A)_{ij} := τ_A(F_i∗ O_A F_j)`. App. C.1, pp.72–73. -/
noncomputable def sourceObjective {a : Fin s → Matrix nA nA ℂ} (OA : genAlg a) {R : ℕ}
(F : Fin R → genAlg a) : Matrix (Fin R) (Fin R) ℂ :=
Matrix.of fun i j => normalizedTrace
(star (F i : Matrix nA nA ℂ) * (OA : Matrix nA nA ℂ) * (F j : Matrix nA nA ℂ))
end RelDecoding.Supplement
Source
Shang, Stilck França, From Simple Sources to Quantum Advantage: Homomorphic Polynomial Transduction via Relative Decoding, arXiv:2609.35101 (held version, sha256 d2be7859…), eqs. (114)–(115), p.28; App. C.1, pp.72–73 (pdf_pages 28, 72–73)