Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma 31

Open
QAlgorithms.Papers.Independent.Intakenext100.P56.cmi_classical_conditioning

by Gohan · Oct 31, 2026 · Mathlib 0df444a (Lean v4.33.1)

prove2quantum

All systems are finite-dimensional complex Hilbert spaces with nonempty finite bases. States are positive semidefinite operators of trace one; pure kets have norm one. All marginal states below are partial traces of the stated joint state. Logarithms are natural, and spectral powers vanish on the kernel, including at exponent zero.

Let pxp_xpx​ be a finite law and ρABCx\rho^x_{ABC}ρABCx​ be states. For ρABCX=∑xpxρABCx⊗∣x⟩⟨x∣X\rho_{ABCX}=\sum_xp_x\rho^x_{ABC}\otimes|x\rangle\langle x|_XρABCX​=∑x​px​ρABCx​⊗∣x⟩⟨x∣X​ and α>0\alpha>0α>0, α≠1\alpha\ne1α=1,

Iα(A;B∣CX)ρ=αα−1log⁡∑xpxexp⁡ ⁣(α−1αIα(A;B∣C)ρx).I_\alpha(A;B|CX)_\rho=\frac{\alpha}{\alpha-1}\log\sum_xp_x\exp\!\left(\frac{\alpha-1}{\alpha}I_\alpha(A;B|C)_{\rho^x}\right).Iα​(A;B∣CX)ρ​=α−1α​logx∑​px​exp(αα−1​Iα​(A;B∣C)ρx​).

At the continuous endpoint, I1(A;B∣CX)ρ=∑xpxI1(A;B∣C)ρxI_1(A;B|CX)_\rho=\sum_xp_xI_1(A;B|C)_{\rho^x}I1​(A;B∣CX)ρ​=∑x​px​I1​(A;B∣C)ρx​. The exponential uses base eee, as required by the paper's natural logarithms. The identity describes classical conditioning.

Preamble
import Definitions.Def_QAlgorithms_Papers_Independent_Intakenext100_P56
Formal statement
noncomputable section

open Filter Topology

open scoped Matrix ComplexConjugate ComplexOrder

namespace QAlgorithms.Papers.Independent.Intakenext100.P56

attribute [local instance] Classical.propDecidable Classical.decEq

open QAlgorithms

open QAlgorithms.Papers.Independent.Information100.P56 (IsLaw entropy average holevo)

open QAlgorithms.Papers.Independent.Information100.P71 (supportPower)

open QAlgorithms.Papers.Independent.Dependency50.P21 (channelAdjoint petz extendedLog marginalAB)

open QAlgorithms.Papers.Independent.Information100.P21 (Op FiniteLOCC)

open QAlgorithms.Papers.Independent.Dependency50.P35 (marginalAC)

open QAlgorithms.Papers.Independent.Dependency50.P07 (bipartiteProduct)

open QAlgorithms.Papers.Independent.Information100.P67 (squaredFidelity)

open QAlgorithms.Papers.Independent.Advance50.P25 (renyiEntropyNats)

variable {A B C D E X : Type}

variable [Fintype A] [Fintype B] [Fintype C] [Fintype D] [Fintype E] [Fintype X]

variable [Nonempty A] [Nonempty B] [Nonempty C] [Nonempty D] [Nonempty E] [Nonempty X]

open scoped Matrix.Norms.L2Operator

/-- arXiv:1410.1443v3, PDF pp.33–34, Lemma 31, (A.12).
The removable one endpoint is stated separately, never by totalized division. -/
theorem cmi_classical_conditioning (a : ℝ) (ha : 0 < a)
    (p : X → ℝ) (hp : IsLaw p) (rho : X → Op (A × (B × C)))
    (hrho : ∀ x, IsDensityMatrix (rho x)) :
    (a ≠ 1 → renyiCMI a (conditionedFlag p rho) =
      a/(a-1) * Real.log (∑ x, p x * Real.exp ((a-1)/a * renyiCMI a (rho x)))) ∧
    renyiCMI 1 (conditionedFlag p rho) = ∑ x, p x * renyiCMI 1 (rho x) := by sorry

end QAlgorithms.Papers.Independent.Intakenext100.P56

end
Source
Rényi squashed entanglement, discord, and relative entropy differences, arXiv:1410.1443v3, PDF p. 33, Lemma 31
Read-back

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

Let A,B,C,XA,B,C,XA,B,C,X be finite index types with decidable equality, let a>0a>0a>0, let p:X→Rp:X\to\mathbb Rp:X→R satisfy px≥0p_x\ge0px​≥0 and ∑xpx=1\sum_xp_x=1∑x​px​=1, and let every ρx\rho_xρx​ be a positive semidefinite complex matrix of trace 111 on A×(B×C)A\times(B\times C)A×(B×C). Write H(Z)=−Re⁡tr⁡(Zlog⁡Z)H(Z)=-\operatorname{Re}\operatorname{tr}(Z\log Z)H(Z)=−Retr(ZlogZ), using continuous functional calculus and the total real logarithm. For a matrix ZZZ, write Z[t]Z^{[t]}Z[t] for functional calculus with the scalar function that is 000 at 000 and utu^tut otherwise; in particular, Z[0]Z^{[0]}Z[0] uses value 000 on the zero eigenspace. For a matrix ω\omegaω on A×(B×C)A\times(B\times C)A×(B×C), let ωAC,ωBC,ωC\omega_{AC},\omega_{BC},\omega_CωAC​,ωBC​,ωC​ denote its indicated partial traces, and define

I1(ω)=H(ωAC)+H(ωBC)−H(ωC)−H(ω),It(ω)=tt−1log⁡Re⁡tr⁡[(StTr⁡A(Rtω[t]Rt)St)[1/t]](t≠1),I_1(\omega)=H(\omega_{AC})+H(\omega_{BC})-H(\omega_C)-H(\omega), \qquad I_t(\omega)=\frac{t}{t-1}\log\operatorname{Re}\operatorname{tr} \left[\left(S_t\operatorname{Tr}_A(R_t\omega^{[t]}R_t)S_t\right)^{[1/t]}\right] \quad(t\ne1),I1​(ω)=H(ωAC​)+H(ωBC​)−H(ωC​)−H(ω),It​(ω)=t−1t​logRetr[(St​TrA​(Rt​ω[t]Rt​)St​)[1/t]](t=1),

where RtR_tRt​ acts as ωAC[(1−t)/2]\omega_{AC}^{[(1-t)/2]}ωAC[(1−t)/2]​ on ACACAC and as the identity on BBB, and St=IB⊗ωC[(t−1)/2]S_t=I_B\otimes\omega_C^{[(t-1)/2]}St​=IB​⊗ωC[(t−1)/2]​. Form the flagged matrix Ω\OmegaΩ on A×(B×(C×X))A\times(B\times(C\times X))A×(B×(C×X)) by

Ω(u,(v,(w,x))),(u′,(v′,(w′,y)))={px(ρx)(u,(v,w)),(u′,(v′,w′)),x=y,0,x≠y.\Omega_{(u,(v,(w,x))),(u',(v',(w',y)))} =\begin{cases} p_x(\rho_x)_{(u,(v,w)),(u',(v',w'))},&x=y,\\ 0,&x\ne y. \end{cases}Ω(u,(v,(w,x))),(u′,(v′,(w′,y)))​={px​(ρx​)(u,(v,w)),(u′,(v′,w′))​,0,​x=y,x=y.​

Then both assertions hold:

a≠1 ⟹ Ia(Ω)=aa−1log⁡ ⁣(∑xpxexp⁡ ⁣(a−1aIa(ρx))),I1(Ω)=∑xpxI1(ρx).a\ne1\ \Longrightarrow\ I_a(\Omega)=\frac{a}{a-1}\log\!\left(\sum_xp_x \exp\!\left(\frac{a-1}{a}I_a(\rho_x)\right)\right), \qquad I_1(\Omega)=\sum_xp_xI_1(\rho_x).a=1 ⟹ Ia​(Ω)=a−1a​log(x∑​px​exp(aa−1​Ia​(ρx​))),I1​(Ω)=x∑​px​I1​(ρx​).

Zero probabilities are allowed, and density matrices are required even at those labels. The probability-law hypothesis cannot hold when XXX is empty; together with the density hypotheses it also excludes empty subsystem types.

Human review
  • Endorsed by Shuze Chen · Oct 31, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Gohan · Oct 31, 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