Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem II.18: For empirical P and R<I(P;W), codes have |M|≥(n+1)^(d−ad)2^(nR), e≤8(n+1)^(ad)2^(−nEg(R,P)), if n≥n0(a,d,P).

Open
QAlgorithms.Papers.Independent.Renew46.P43.constantCompositionCoding

by Goku · Oct 22, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

Fix a finite stationary channel WWW with input alphabet size aaa and output dimension ddd. Let PPP be a realizable length-nnn empirical law, n>0n>0n>0 and R<I(P;W)R<I(P;W)R<I(P;W). Here Eg(R,P)=sup⁡R≤L′−Lmin⁡{μi(L,P),12μc(L′,P)}E_g(R,P)=\sup_{R\le L'-L}\min\{\mu_i(L,P),\tfrac12\mu_c(L',P)\}Eg​(R,P)=supR≤L′−L​min{μi​(L,P),21​μc​(L′,P)}, where μi=inf⁡H(V∣P)>LD(V∥W∣P)\mu_i=\inf_{H(V\mid P)>L}D(V\Vert W\mid P)μi​=infH(V∣P)>L​D(V∥W∣P) and μc=inf⁡H(ν)≤L′D(ν∥PW)\mu_c=\inf_{H(\nu)\le L'}D(\nu\Vert PW)μc​=infH(ν)≤L′​D(ν∥PW). For fixed a,d,Pa,d,Pa,d,P, one threshold n0(a,d,P)n_0(a,d,P)n0​(a,d,P) works for every channel WWW, every realizable positive length nnn, and every R<I(P;W)R<I(P;W)R<I(P;W).

∣M∣≥(n+1)d−ad2nR,e≤8(n+1)ad2−nEg(R,P)(n≥n0(a,d,P), R<I(P;W)).|M|\ge(n+1)^{d-ad}2^{nR},\quad e\le8(n+1)^{ad}2^{-nE_g(R,P)}\quad(n\ge n_0(a,d,P),\ R<I(P;W)).∣M∣≥(n+1)d−ad2nR,e≤8(n+1)ad2−nEg​(R,P)(n≥n0​(a,d,P), R<I(P;W)).

This is the source result with its stated operational or information-theoretic domain.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Renew46_P43
Formal statement
noncomputable section

namespace QAlgorithms.Papers.Independent.Renew46.P43

open scoped BigOperators Matrix ComplexConjugate ComplexOrder ENNReal Matrix.Norms.L2Operator

open Filter MeasureTheory

open QAlgorithms.Papers.Independent.Information100.P08 (IsPure)

open QAlgorithms.Papers.Independent.Information100.P06 (Below)

open QAlgorithms.Papers.Independent.Information100.P60 (copies relativeBits)

open QAlgorithms.Papers.Independent.Dependency50.P19 (count both)

open QAlgorithms.Papers.Independent.Information100.P22 (coherentInformation)

attribute [local instance] Classical.propDecidable Classical.decEq

open QAlgorithms.Papers.Independent.Claimed50.P40 (InBlocks)

attribute [local instance] QAlgorithms.Papers.Independent.Renew46.P43.matrixMeasurableSpace

/-- Theorem II.18, PDF pp.34-35: retain the negative polynomial prefactor and
factor 8. The printed n0(a,d,P) precedes the channel W, block length and rate R.
The proof's small-error transition caveat is recorded without weakening this
fully pinned theorem. -/
theorem constantCompositionCoding {a d : ℕ} (ha : 0 < a) (hd : 0 < d)
    (P : PMF (Fin a)) :
    ∃ n0 : ℕ, ∀ (W : FiniteChannel a (fullAlgebra d hd)) (n : ℕ), 0 < n → n0 ≤ n →
      (∃ w : Fin n → Fin a, ∀ x, empiricalWeight w x = (P x).toReal) →
      ∀ R : ℝ, R < W.information P →
      ∃ c : FiniteCode W n, c.Composition P ∧
        (c.messages : ℝ) ≥ (n+1 : ℝ)^((d : ℤ) - (a : ℤ)*d) * exp2 (n * R) ∧
        c.error ≤ 8 * (n+1 : ℝ)^(a*d) * divergenceTail n (greedyExponent W R P) := by sorry

end QAlgorithms.Papers.Independent.Renew46.P43

end
Source
Coding Theorems of Quantum Information Theory, arXiv:quant-ph/9907077v1, PDF p. 34, Theorem II.18
Read-back

What the Lean code literally says, in plain math · gpt-6.1-sol

For every pair of positive natural numbers a,da,da,d and every probability mass function PPP on the alphabet {0,…,a−1}\{0,\ldots,a-1\}{0,…,a−1}, there exists a natural number n0n_0n0​ such that the following holds uniformly for every channel WWW assigning a positive semidefinite complex d×dd\times dd×d matrix WxW_xWx​ of trace 111 to each alphabet symbol xxx, every positive natural number n≥n0n\ge n_0n≥n0​, and every real R<IW(P)R<I_W(P)R<IW​(P), provided there exists a length-nnn word in which each symbol xxx occurs exactly nP(x)nP(x)nP(x) times. Here IW(P)=H(∑xP(x)Wx)−∑xP(x)H(Wx)I_W(P)=H(\sum_xP(x)W_x)-\sum_xP(x)H(W_x)IW​(P)=H(∑x​P(x)Wx​)−∑x​P(x)H(Wx​), where H(ρ)=−Re⁡Tr⁡[f(ρ)]H(\rho)=-\operatorname{Re}\operatorname{Tr}[f(\rho)]H(ρ)=−ReTr[f(ρ)], f(t)=tlog⁡2tf(t)=t\log_2tf(t)=tlog2​t, with f(ρ)f(\rho)f(ρ) defined by continuous functional calculus. There then exists a code with a positive number MMM of messages, an encoding word wmw_mwm​ of length nnn for each message mmm, and positive semidefinite decoding matrices D⊥,D0,…,DM−1D_\bot,D_0,\ldots,D_{M-1}D⊥​,D0​,…,DM−1​ on (Cd)⊗n(\mathbb C^d)^{\otimes n}(Cd)⊗n whose sum is the identity, such that every wmw_mwm​ has empirical frequencies PPP and

M≥(n+1) d−ad 2nR,e(c)≤8(n+1)ad Tn(GW(R,P)).M\ge(n+1)^{\,d-ad}\,2^{nR}, \qquad e(c)\le 8(n+1)^{ad}\,T_n(G_W(R,P)).M≥(n+1)d−ad2nR,e(c)≤8(n+1)adTn​(GW​(R,P)).

The error is precisely e(c)=max⁡m{1−Re⁡Tr⁡[(⨂i=0n−1Wwm(i))Dm]}e(c)=\max_m\{1-\operatorname{Re}\operatorname{Tr}[(\bigotimes_{i=0}^{n-1}W_{w_m(i)})D_m]\}e(c)=maxm​{1−ReTr[(⨂i=0n−1​Wwm​(i)​)Dm​]}; the outcome ⊥\bot⊥ is an additional decoding outcome. The tail function is Tn(+∞)=0T_n(+\infty)=0Tn​(+∞)=0 and Tn(D)=2−nDT_n(D)=2^{-nD}Tn​(D)=2−nD for finite DDD. Its argument is

GW(R,P)=sup⁡L,L′∈RR≤L′−Lmin⁡ ⁣{inf⁡Vx density matricesL<∑xP(x)H(Vx)∑xP(x) D(Vx∥Wx),12inf⁡ρ a density matrixH(ρ)≤L′D ⁣(ρ∥∑xP(x)Wx)}.G_W(R,P)= \sup_{\substack{L,L'\in\mathbb R\\R\le L'-L}} \min\!\left\{ \inf_{\substack{V_x\text{ density matrices}\\L<\sum_xP(x)H(V_x)}} \sum_xP(x)\,\mathcal D(V_x\Vert W_x), \quad \frac12\inf_{\substack{\rho\text{ a density matrix}\\H(\rho)\le L'}} \mathcal D\!\left(\rho\middle\Vert\sum_xP(x)W_x\right) \right\}.GW​(R,P)=L,L′∈RR≤L′−L​sup​min⎩⎨⎧​Vx​ density matricesL<∑x​P(x)H(Vx​)​inf​x∑​P(x)D(Vx​∥Wx​),21​ρ a density matrixH(ρ)≤L′​inf​D(ρ​x∑​P(x)Wx​)⎭⎬⎫​.

All matrices Vx,ρV_x,\rhoVx​,ρ in these infima are d×dd\times dd×d. For density matrices σ,τ\sigma,\tauσ,τ, D(σ∥τ)=+∞\mathcal D(\sigma\Vert\tau)=+\inftyD(σ∥τ)=+∞ when ker⁡τ⊈ker⁡σ\ker\tau\not\subseteq\ker\sigmakerτ⊆kerσ; otherwise it is max⁡{Re⁡Tr⁡[σln⁡σ−σln⁡τ],0}/ln⁡2\max\{\operatorname{Re}\operatorname{Tr}[\sigma\ln\sigma-\sigma\ln\tau],0\}/\ln2max{ReTr[σlnσ−σlnτ],0}/ln2, with the matrix functions interpreted by the functional calculus in the code. These suprema and infima take values in [0,+∞][0,+\infty][0,+∞], so an empty infimum is +∞+\infty+∞. No positivity assumption is imposed on every P(x)P(x)P(x), and RRR need not be nonnegative. The exact-composition premise can fail—for example, if some nP(x)nP(x)nP(x) is not an integer—in which case the implication asserts no code existence for that nnn.

Human review
  • Endorsed by Shuze Chen · Oct 22, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Goku · Oct 22, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me