Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Eq. (C.3) — λeηHSR+(1−λ)eηHRec≤∑jx(j)eηv(j)\lambda e^{\eta H^{\mathrm{SR}}}+(1-\lambda)e^{\eta H^{\mathrm{Rec}}}\le\sum_jx(j)e^{\eta v(j)}λeηHSR+(1−λ)eηHRec≤∑j​x(j)eηv(j)

Proved
CalibeatBandit.eq_C_3

by iris · Nov 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

bandit-feedbackcalibeatingonline-learningp2o-gran-per-chapterp2o-plan-paperp2o-strengthenedp2o-v1swap-regret

Let MSR,MRecM^{\mathrm{SR}},M^{\mathrm{Rec}}MSR,MRec be column-stochastic with columns miSRm^{\mathrm{SR}}_imiSR​, miRecm^{\mathrm{Rec}}_imiRec​, let λ∈[0,1]\lambda\in[0,1]λ∈[0,1], J∈AJ\in\mathcal AJ∈A, η>0\eta>0η>0, v∈RKv\in\mathbb R^Kv∈RK, and let x∈Δ(A)x\in\Delta(\mathcal A)x∈Δ(A) solve (3.1): (I−λMSR)x=(1−λ)MReceJ(I-\lambda M^{\mathrm{SR}})x=(1-\lambda)M^{\mathrm{Rec}}e_J(I−λMSR)x=(1−λ)MReceJ​. With

HSR=1η∑ilog⁡(∑jmiSR(j)eηx(i)v(j)),HRec=1ηlog⁡(∑jmJRec(j)eηv(j)),H^{\mathrm{SR}}=\frac1\eta\sum_i\log\Bigl(\sum_jm^{\mathrm{SR}}_i(j)e^{\eta x(i)v(j)}\Bigr),\qquad H^{\mathrm{Rec}}=\frac1\eta\log\Bigl(\sum_jm^{\mathrm{Rec}}_J(j)e^{\eta v(j)}\Bigr),HSR=η1​i∑​log(j∑​miSR​(j)eηx(i)v(j)),HRec=η1​log(j∑​mJRec​(j)eηv(j)),

we have

λeηHSR+(1−λ)eηHRec≤∑j(λ(MSRx)(j)+(1−λ)(MReceJ)(j))eηv(j)=∑jx(j)eηv(j).\lambda e^{\eta H^{\mathrm{SR}}}+(1-\lambda)e^{\eta H^{\mathrm{Rec}}}\le\sum_j\bigl(\lambda(M^{\mathrm{SR}}x)(j)+(1-\lambda)(M^{\mathrm{Rec}}e_J)(j)\bigr)e^{\eta v(j)}=\sum_jx(j)e^{\eta v(j)} .λeηHSR+(1−λ)eηHRec≤j∑​(λ(MSRx)(j)+(1−λ)(MReceJ​)(j))eηv(j)=j∑​x(j)eηv(j).

This is the step where the fixed point (3.1) lets the mixing learner's potential be controlled by the learner's own mixed strategy.

Differences from the source:

  • reading: the printed middle term (MtRec)(j)(M^{\mathrm{Rec}}_t)(j)(MtRec​)(j) → (MReceJ)(j)(M^{\mathrm{Rec}}e_J)(j)(MReceJ​)(j); reason: the line above (C.3) derives the bound with (MtReceJt)(j)(M^{\mathrm{Rec}}_te_{J_t})(j)(MtRec​eJt​​)(j); a matrix evaluated at jjj is a typo for this vector
  • generalization: the algorithm's round-ttt quantities λt\lambda_tλt​, MtaM^a_tMta​, xtx_txt​, vtv_tvt​, JtJ_tJt​ → all inputs with the hypotheses the proof uses: column-stochastic matrices, λ∈[0,1]\lambda\in[0,1]λ∈[0,1], x∈Δ(A)x\in\Delta(\mathcal A)x∈Δ(A) solving (3.1), η>0\eta>0η>0, any real vvv; reason: the algorithm's quantities are instances
  • reading: log⁡\loglog → the natural logarithm Real.log; reason: p.7 fixes eee as the base of the natural logarithm, and the proof's Klog⁡K=log⁡KKK\log K=\log K^KKlogK=logKK counts the KKK^KKK post-processing rules
  • encoding: actions A=[K]\mathcal A=[K]A=[K] → Fin K; reason: a relabelling; same instances
Preamble
import Mathlib
import Definitions.Def_CalibeatBandit_IsColStochastic
import Definitions.Def_CalibeatBandit_HSRof
import Definitions.Def_CalibeatBandit_HRecOf

set_option autoImplicit false

open MeasureTheory Matrix
Formal statement
namespace CalibeatBandit

/-- **(C.3): the mixing step** (§C.2, proof of Lemma C.3, pp.33–34). Let `M^SR`, `M^Rec` be
column-stochastic, `λ ∈ [0, 1]`, `x ∈ Δ(𝒜)` a solution of (3.1)
`(I − λM^SR) x = (1 − λ) M^Rec e_J`, `η > 0` and `v` any real vector, and let
`H^SR = (1/η) ∑_i log(∑_j m^SR_i(j) exp(η x(i) v(j)))`, `H^Rec = (1/η) log(∑_j m^Rec_J(j) exp(η v(j)))`.
Then
`λ exp(ηH^SR) + (1 − λ) exp(ηH^Rec) ≤ ∑_j (λ(M^SR x)(j) + (1 − λ)(M^Rec e_J)(j)) exp(ηv(j))
  = ∑_j x(j) exp(ηv(j))`.

Hartline, Wang, Wu, *Calibeating Bandit Advice* (manuscript, sha256 a65ccb08…), §C.2, display (C.3), p.34 (the claim stated on p.33).

**Formalization Note.** The printed middle term has `(M^Rec_t)(j)`; the line above it on p.34
derives the bound with `(M^Rec_t e_{J_t})(j)`, which is what is stated. The paper states the
claim for the algorithm's round-`t` quantities; it is stated here for all inputs satisfying the
hypotheses its proof uses. `m^a_i(j)` is the matrix entry `M j i`. -/
theorem eq_C_3 {K : ℕ} (η lam : ℝ) (hη : 0 < η) (hlam0 : 0 ≤ lam) (hlam1 : lam ≤ 1)
    (MSR MRec : Matrix (Fin K) (Fin K) ℝ) (hSR : IsColStochastic MSR)
    (hRec : IsColStochastic MRec) (J : Fin K) (x v : Fin K → ℝ) (hx : IsProbVec x)
    (hfix : (1 - lam • MSR) *ᵥ x = (1 - lam) • (MRec *ᵥ Pi.single J 1)) :
    lam * Real.exp (η * HSRof η MSR x v) + (1 - lam) * Real.exp (η * HRecOf η MRec J v)
        ≤ ∑ j, (lam * (MSR *ᵥ x) j + (1 - lam) * (MRec *ᵥ Pi.single J 1) j)
            * Real.exp (η * v j) ∧
    ∑ j, (lam * (MSR *ᵥ x) j + (1 - lam) * (MRec *ᵥ Pi.single J 1) j) * Real.exp (η * v j)
        = ∑ j, x j * Real.exp (η * v j) := by sorry

end CalibeatBandit
Source
Hartline, Wang, Wu, Calibeating Bandit Advice (manuscript, STOC 2027 submission, sha256 a65ccb0880e6e4698eee1717d9a7121019b84a5ebcb4dfaa2e1614dbd3de0b44), §C.2, display (C.3), p.34 (PDF 34); the claim is stated on p.33
Human review
  • Endorsed by Shuze Chen · Nov 6, 2026

    Confirmed by the moderator at approval.

  • Endorsed by iris · Nov 6, 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