Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

§C.2 (p.32) — three-part decomposition of RTSR,pR^{\mathrm{SR},p}_TRTSR,p​ and RTRec,pR^{\mathrm{Rec},p}_TRTRec,p​

Proved
CalibeatBandit.decomposition_C_2

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

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

Let 0≤μ≤10\le\mu\le10≤μ≤1, let xt∈Δ(A)x_t\in\Delta(\mathcal A)xt​∈Δ(A), yt∈[0,1]Ky_t\in[0,1]^Kyt​∈[0,1]K, vt∈RKv_t\in\mathbb R^Kvt​∈RK and Jt∈AJ_t\in\mathcal AJt​∈A for t=1,…,Tt=1,\dots,Tt=1,…,T, and let pt=(1−μ)xt+μup_t=(1-\mu)x_t+\mu upt​=(1−μ)xt​+μu with u=1K1u=\frac1K\mathbf 1u=K1​1. Then

RTSR,p≤μT+max⁡f∑t=1T(∑ixt(i)vt(f(i))−⟨xt,vt⟩)+max⁡f∑t=1T∑ixt(i)(yt(f(i))−vt(f(i)))+∑t=1T⟨xt,vt−yt⟩R^{\mathrm{SR},p}_T\le\mu T+\max_f\sum_{t=1}^T\Bigl(\sum_ix_t(i)v_t(f(i))-\langle x_t,v_t\rangle\Bigr)+\max_f\sum_{t=1}^T\sum_ix_t(i)\bigl(y_t(f(i))-v_t(f(i))\bigr)+\sum_{t=1}^T\langle x_t,v_t-y_t\rangleRTSR,p​≤μT+fmax​t=1∑T​(i∑​xt​(i)vt​(f(i))−⟨xt​,vt​⟩)+fmax​t=1∑T​i∑​xt​(i)(yt​(f(i))−vt​(f(i)))+t=1∑T​⟨xt​,vt​−yt​⟩

and

RTRec,p≤μT+max⁡f∑t=1T(vt(f(Jt))−⟨xt,vt⟩)+max⁡f∑t=1T(yt(f(Jt))−vt(f(Jt)))+∑t=1T⟨xt,vt−yt⟩,R^{\mathrm{Rec},p}_T\le\mu T+\max_f\sum_{t=1}^T\bigl(v_t(f(J_t))-\langle x_t,v_t\rangle\bigr)+\max_f\sum_{t=1}^T\bigl(y_t(f(J_t))-v_t(f(J_t))\bigr)+\sum_{t=1}^T\langle x_t,v_t-y_t\rangle ,RTRec,p​≤μT+fmax​t=1∑T​(vt​(f(Jt​))−⟨xt​,vt​⟩)+fmax​t=1∑T​(yt​(f(Jt​))−vt​(f(Jt​)))+t=1∑T​⟨xt​,vt​−yt​⟩,

where fff ranges over all maps A→A\mathcal A\to\mathcal AA→A.

The four terms are the exploration loss, the regret on corrected rewards (Lemma C.3), the estimation loss (Lemma C.4) and the correction loss (Lemma C.5).

Formalization Note The inequalities are deterministic; ptp_tpt​ is constrained only on the rounds 0,…,T−10,\dots,T-10,…,T−1.

Differences from the source:

  • generalization: the algorithm's xtx_txt​, vtv_tvt​, JtJ_tJt​ with 0<μ≤1/20<\mu\le1/20<μ≤1/2 → arbitrary xt∈Δ(A)x_t\in\Delta(\mathcal A)xt​∈Δ(A), vt∈RKv_t\in\mathbb R^Kvt​∈RK, JtJ_tJt​ and 0≤μ≤10\le\mu\le10≤μ≤1; reason: the algorithm's quantities satisfy these hypotheses
  • encoding: rounds t=1,…,Tt=1,\dots,Tt=1,…,T, actions A=[K]\mathcal A=[K]A=[K] → rounds t=0,…,T−1t=0,\dots,T-1t=0,…,T−1 (Finset.range T), actions Fin K; reason: a relabelling; same instances
Preamble
import Mathlib
import Definitions.Def_CalibeatBandit_IsProbVec
import Definitions.Def_CalibeatBandit_pseudoSwapRegret
import Definitions.Def_CalibeatBandit_pseudoRewardLoss

set_option autoImplicit false

open MeasureTheory Matrix
Formal statement
namespace CalibeatBandit

/-- **The three-part decomposition of the pseudo regrets** (§C.2, p.32). Let
`p_t = (1 − μ) x_t + μ u` with `x_t ∈ Δ(𝒜)`, `u = (1/K)·𝟏`, `0 ≤ μ ≤ 1`, rewards
`y_t ∈ [0,1]^K`, arbitrary vectors `v_t` and actions `J_t`. Then
`R^{SR,p}_T ≤ μT + max_f ∑_t (∑_i x_t(i) v_t(f(i)) − ⟨x_t, v_t⟩)
  + max_f ∑_t ∑_i x_t(i)(y_t(f(i)) − v_t(f(i))) + ∑_t ⟨x_t, v_t − y_t⟩`
and
`R^{Rec,p}_T ≤ μT + max_f ∑_t (v_t(f(J_t)) − ⟨x_t, v_t⟩) + max_f ∑_t (y_t(f(J_t)) − v_t(f(J_t)))
  + ∑_t ⟨x_t, v_t − y_t⟩`.

Hartline, Wang, Wu, *Calibeating Bandit Advice* (manuscript, sha256 a65ccb08…), §C.2, p.32 (the two displays before Lemma C.3).

**Formalization Note.** Deterministic: the inequalities hold for every realization. The paper
states them for the algorithm's `x_t`, `v_t` with `0 < μ ≤ 1/2`; here `x`, `v`, `J` are arbitrary
(with `x_t ∈ Δ(𝒜)`) and `0 ≤ μ ≤ 1`. Rounds are `0, …, T − 1`; `p` is only constrained on
those rounds. -/
theorem decomposition_C_2 {K T : ℕ} (μ : ℝ) (hμ0 : 0 ≤ μ) (hμ1 : μ ≤ 1)
    (x v y p : ℕ → Fin K → ℝ) (J : ℕ → Fin K) (hx : ∀ t < T, IsProbVec (x t))
    (hy : ∀ t < T, ∀ j, y t j ∈ Set.Icc (0 : ℝ) 1)
    (hp : ∀ t < T, ∀ j, p t j = (1 - μ) * x t j + μ / K) :
    pseudoSwapRegret T p y ≤ μ * T
        + Finset.univ.sup' ⟨id, Finset.mem_univ _⟩ (fun f : Fin K → Fin K =>
            ∑ t ∈ Finset.range T, (∑ i, x t i * v t (f i) - x t ⬝ᵥ v t))
        + Finset.univ.sup' ⟨id, Finset.mem_univ _⟩ (fun f : Fin K → Fin K =>
            ∑ t ∈ Finset.range T, ∑ i, x t i * (y t (f i) - v t (f i)))
        + ∑ t ∈ Finset.range T, x t ⬝ᵥ (v t - y t) ∧
    pseudoRewardLoss T p y J ≤ μ * T
        + Finset.univ.sup' ⟨id, Finset.mem_univ _⟩ (fun f : Fin K → Fin K =>
            ∑ t ∈ Finset.range T, (v t (f (J t)) - x t ⬝ᵥ v t))
        + Finset.univ.sup' ⟨id, Finset.mem_univ _⟩ (fun f : Fin K → Fin K =>
            ∑ t ∈ Finset.range T, (y t (f (J t)) - v t (f (J t))))
        + ∑ t ∈ Finset.range T, x t ⬝ᵥ (v t - y t) := by sorry

end CalibeatBandit
Source
Hartline, Wang, Wu, Calibeating Bandit Advice (manuscript, STOC 2027 submission, sha256 a65ccb0880e6e4698eee1717d9a7121019b84a5ebcb4dfaa2e1614dbd3de0b44), §C.2, the two displays before Lemma C.3, p.32 (PDF 32)
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