The least rank- majorant
DefinitionSpectralFeedback_Supplement_leastMajorantlinear-algebramatrix-analysisoptimizationp2o-gran-per-chapterp2o-plan-paperp2o-v1quantum-information
Let be a spectrum and . Choose the smallest with
and put , with entries equal to . For a probability vector this is the least rank- probability vector majorizing . The unlimited-message value is , and Theorems C.4 and E.1 characterize when finitely many symbols attain it.
Formalization Note Three definitions: wfLevel (the level ), wfIndex (the head size ) and leastMajorant (). Indices are 0-based: lam i is , so becomes the sum over , and entry of is lam i for , for and after that. For a nonnegative decreasing list with the index always qualifies. The fallback value is used only when no index qualifies, which no statement of the mission uses.
Definition code
import Mathlib
namespace SpectralFeedback
namespace Supplement
/-- **The water level `τ_j`** of the least rank-`r′` majorant (Qi, *Spectral Structure and Exact
Arithmetic of Finite-Feedback Measurement Discrimination*, manuscript (Nov. 2026), §C.4, p.29
(PDF p.31); §E.1, p.75 (PDF p.77)): `τ = ∑_{i>j} λ_i / (r′ − j)`, the sum running over
`j < i ≤ n`.
**Formalization Note** 0-based: `lam i` is the paper's `λ_{i+1}`, so the sum is over
`i ∈ [j, n)`. Used only with `j < r′`, where the denominator is positive. -/
noncomputable def wfLevel (lam : ℕ → ℝ) (n r' j : ℕ) : ℝ :=
(∑ i ∈ Finset.Ico j n, lam i) / ((r' : ℝ) - (j : ℝ))
open Classical in
/-- **The head size `j`** of the least rank-`r′` majorant (Qi, manuscript (Nov. 2026), §C.4, p.29
(PDF p.31); §E.1, p.75): the smallest `j ∈ {0, …, r′ − 1}` with `λ_{j+1} ≤ τ_j`.
**Formalization Note** 0-based (`lam j` is `λ_{j+1}`). For a nonnegative decreasing list with
`1 ≤ r′ ≤ n` the index `j = r′ − 1` always qualifies, so the minimum exists; the fallback value `0`
is used only when no index qualifies (e.g. `r′ = 0`). -/
noncomputable def wfIndex (lam : ℕ → ℝ) (n r' : ℕ) : ℕ :=
if h : ∃ j, j < r' ∧ lam j ≤ wfLevel lam n r' j then Nat.find h else 0
/-- **The least rank-`r′` probability vector majorizing `λ`** (Qi, *Spectral Structure and Exact
Arithmetic of Finite-Feedback Measurement Discrimination*, manuscript (Nov. 2026), §C.4, p.29
(PDF p.31); §E.1, p.75 (PDF p.77)): with `j = wfIndex` and `τ = τ_j`,
`q = (λ_1, …, λ_j, τ, …, τ, 0, …, 0) ∈ ℝ^n`, with `r′ − j` entries equal to `τ`.
**Formalization Note** 0-based: entry `i` is `lam i` for `i < j`, `τ` for `j ≤ i < r′`, and `0`
for `r′ ≤ i < n`. -/
noncomputable def leastMajorant (lam : ℕ → ℝ) (n r' : ℕ) : Fin n → ℝ :=
fun i => if (i : ℕ) < wfIndex lam n r' then lam i
else if (i : ℕ) < r' then wfLevel lam n r' (wfIndex lam n r') else 0
end Supplement
end SpectralFeedback
Source
Qi, Spectral Structure and Exact Arithmetic of Finite-Feedback Measurement Discrimination, manuscript (Nov. 2026; PDF sha256 a5b5cf40…eabde9cf), §C.4, p.29 (PDF p.31); §E.1, p.75 (PDF p.77)