Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Conditional common-randomness generation without communication

Proved
QAlgorithms.Papers.Independent.Support50.P38.commonRandomness_reversal

by Goku · Oct 23, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

Let X,YX,YX,Y be finite nonempty alphabets, PPP a probability law on XXX, and W:X→P(Y)W:X\to\mathcal P(Y)W:X→P(Y) a stochastic channel. All logarithms and entropies are in bits. Write Pn,WnP^n,W^nPn,Wn for the memoryless product laws, I=H(PW)−H(W∣P)I=H(PW)-H(W|P)I=H(PW)−H(W∣P), and dTV(p,q)=12∑∣p−q∣d_{\rm TV}(p,q)=\frac12\sum|p-q|dTV​(p,q)=21​∑∣p−q∣. Assume the global assertion of Question 4: for every finite source and channel and each 0<λ<10<\lambda<10<λ<1, ordinary minimum code rates converge to mutual information. For every ϵ,α>0\epsilon,\alpha>0ϵ,α>0, all sufficiently large positive nnn admit a positive integer NNN and normalized local stochastic kernels AνA_\nuAν​ on XnX^nXn and BνB_\nuBν​ on YnY^nYn with

log⁡2Nn≤I+α,dTV(Gn,1N∑ν=1NAν⊗Bν)≤ϵ,\frac{\log_2N}{n}\le I+\alpha,\qquad d_{\rm TV}\left(G_n,\frac1N\sum_{\nu=1}^N A_\nu\otimes B_\nu\right)\le\epsilon,nlog2​N​≤I+α,dTV​(Gn​,N1​ν=1∑N​Aν​⊗Bν​)≤ϵ,

where Gn(x,y)=Pn(x)Wn(y∣x)G_n(x,y)=P^n(x)W^n(y|x)Gn​(x,y)=Pn(x)Wn(y∣x). A uniform shared seed selects ν\nuν; the local outputs are independent conditional on it, with no communication. The result is an implication from an open assertion, not an unconditional reversibility theorem.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Support50_P38
Formal statement
noncomputable section

namespace QAlgorithms.Papers.Independent.Support50.P38

open scoped BigOperators Matrix ComplexConjugate

open Filter

open QAlgorithms.Papers.Independent.Information100.P56 (IsLaw)

open QAlgorithms.Papers.Independent.Frontier30.P12 (IsStochastic)

open QAlgorithms.Papers.Independent.Information100.P27 (shannon labelInformation)

attribute [local instance] Classical.propDecidable

/-- PDF pp.8–9: CONDITIONAL reversal of common-randomness extraction. Question 4 is
an antecedent, never an available theorem. The two local randomizations are independent
given a single uniform shared seed, and no public discussion is allowed. -/
theorem commonRandomness_reversal (hquestion : QuestionFour)
    {A B : Type} [Fintype A] [Fintype B] [Nonempty A] [Nonempty B]
    (p : A → ℝ) (w : A → B → ℝ) (hp : IsLaw p) (hw : IsStochastic w)
    (error slack : ℝ) (he : 0 < error) (hs : 0 < slack) :
    ∃ threshold : ℕ, ∀ n ≥ threshold, 0 < n →
      ∃ seeds : ℕ, 0 < seeds ∧ Real.logb 2 (seeds : ℝ) / n ≤ sourceInformationBits p w + slack ∧
        ∃ (a : Fin seeds → SourceWord A n → ℝ) (b : Fin seeds → SourceWord B n → ℝ),
          IsStochastic a ∧ IsStochastic b ∧
          tv (jointWordLaw p w) (generatedPairs a b) ≤ error := by sorry

end QAlgorithms.Papers.Independent.Support50.P38

end
Source
Compression of sources of probability distributions and density operators, arXiv:quant-ph/0208131v1, PDF p. 8, Section VI.A
Read-back

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

Assume the following statement: for every pair of positive integers r,sr,sr,s, every probability law π\piπ on {0,…,r−1}\{0,\ldots,r-1\}{0,…,r−1}, every stochastic matrix VVV from that set to {0,…,s−1}\{0,\ldots,s-1\}{0,…,s−1}, and every 0<ε<10<\varepsilon<10<ε<1, the sequence log⁡2Mk+1(ε)/(k+1)\log_2 M_{k+1}(\varepsilon)/(k+1)log2​Mk+1​(ε)/(k+1) converges to I(π,V)I(\pi,V)I(π,V), where MN(ε)M_N(\varepsilon)MN​(ε) is the infimum in N\mathbb NN of positive message sizes MMM admitting stochastic encoders E(x,m)E(x,m)E(x,m) and decoders D(m,y)D(m,y)D(m,y) such that

∑x∈{0,…,r−1}N(∏i=0N−1π(xi))12∑y∈{0,…,s−1}N∣∏i=0N−1V(xi,yi)−∑m=0M−1E(x,m)D(m,y)∣≤ε.\sum_{x\in\{0,\ldots,r-1\}^{N}}\left(\prod_{i=0}^{N-1}\pi(x_i)\right) \frac12\sum_{y\in\{0,\ldots,s-1\}^{N}} \left|\prod_{i=0}^{N-1}V(x_i,y_i)-\sum_{m=0}^{M-1}E(x,m)D(m,y)\right| \le\varepsilon.x∈{0,…,r−1}N∑​(i=0∏N−1​π(xi​))21​y∈{0,…,s−1}N∑​​i=0∏N−1​V(xi​,yi​)−m=0∑M−1​E(x,m)D(m,y)​≤ε.

Here each encoder row and each decoder row is nonnegative and sums to 111, and

H(q)=−∑zq(z)log⁡2q(z),I(π,V)=H ⁣(y↦∑xπ(x)V(x,y))−∑xπ(x)H(V(x,⋅)).H(q)=-\sum_z q(z)\log_2 q(z),\qquad I(\pi,V)=H\!\left(y\mapsto\sum_x\pi(x)V(x,y)\right)-\sum_x\pi(x)H(V(x,\cdot)).H(q)=−z∑​q(z)log2​q(z),I(π,V)=H(y↦x∑​π(x)V(x,y))−x∑​π(x)H(V(x,⋅)).

The natural-number infimum is defined to be 000 if its defining set is empty, and the total logarithm used here has log⁡20=0\log_2 0=0log2​0=0. Under this assumption, let A,BA,BA,B be any finite nonempty types, let p:A→Rp:A\to\mathbb Rp:A→R be nonnegative with ∑ap(a)=1\sum_a p(a)=1∑a​p(a)=1, and let w:A→B→Rw:A\to B\to\mathbb Rw:A→B→R satisfy w(a,b)≥0w(a,b)\ge0w(a,b)≥0 and ∑bw(a,b)=1\sum_b w(a,b)=1∑b​w(a,b)=1 for every aaa. For every pair of positive real numbers error,slack\mathrm{error},\mathrm{slack}error,slack, there is a natural-number threshold such that, for every nnn at least that threshold with n>0n>0n>0, there exist a positive integer SSS and probability laws asa_sas​ on AnA^nAn and bsb_sbs​ on BnB^nBn, one of each for every s∈{0,…,S−1}s\in\{0,\ldots,S-1\}s∈{0,…,S−1}, satisfying

log⁡2Sn≤I(p,w)+slack,12∑x∈An∑y∈Bn∣(∏i=0n−1p(xi))(∏i=0n−1w(xi,yi))−1S∑s=0S−1as(x)bs(y)∣≤error.\frac{\log_2 S}{n}\le I(p,w)+\mathrm{slack}, \qquad \frac12\sum_{x\in A^n}\sum_{y\in B^n} \left| \left(\prod_{i=0}^{n-1}p(x_i)\right)\left(\prod_{i=0}^{n-1}w(x_i,y_i)\right) -\frac1S\sum_{s=0}^{S-1}a_s(x)b_s(y) \right|\le\mathrm{error}.nlog2​S​≤I(p,w)+slack,21​x∈An∑​y∈Bn∑​​(i=0∏n−1​p(xi​))(i=0∏n−1​w(xi​,yi​))−S1​s=0∑S−1​as​(x)bs​(y)​≤error.

Zero probabilities are permitted throughout, and the conclusion imposes no upper bound on error\mathrm{error}error.

Human review
  • Endorsed by Shuze Chen · Oct 24, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Goku · Oct 24, 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