Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Corollary 6, p. 13, for BFPL* as on p. 14 (corrected: + E) — Regret_T ≤ Switches_T + E + Σ_e max_i R_e(i)/ε pointwise

Proved
LimitedSwitching.HighProb.corollary_6

by mikedeng1 · Oct 19, 2026 · Mathlib 0df444a (Lean v4.33.1)

follow-the-perturbed-leaderonline-learningp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1regret-boundswitching

Let n≥1n\ge1n≥1, ε>0\varepsilon>0ε>0, any budget S′S'S′ and T≥1T\ge1T≥1, and let (ℓt)(\ell_t)(ℓt​) be a loss sequence with values in [0,1][0,1][0,1]. Let B\mathcal BB be Framework 1 applied to FPLε∗^*_\varepsilonε∗​ (perturbation −Re(i)/ε-R_e(i)/\varepsilon−Re​(i)/ε in epoch eee), run with nonnegative perturbation vectors R0,R1,⋯∈[0,∞)nR_0,R_1,\dots\in[0,\infty)^nR0​,R1​,⋯∈[0,∞)n, one for each epoch. Let EEE be the number of epochs used in rounds 1,…,T1,\dots,T1,…,T. Then, pointwise,

RegretT(B)≤SwitchesT(B)+E+∑e<Emax⁡i∈[n]Re(i)ε.\mathrm{Regret}_T(\mathcal B)\le\mathrm{Switches}_T(\mathcal B)+E+\sum_{e<E}\max_{i\in[n]}\frac{R_e(i)}{\varepsilon}.RegretT​(B)≤SwitchesT​(B)+E+e<E∑​i∈[n]max​εRe​(i)​.

This is the specialisation of Corollary 6 used in the proof of Theorem 8 (p. 14). Combined with Lemma 3, which bounds EEE, and Lemma 9, which bounds the sum of maxima, it gives the high-probability regret bound.

Formalization Note The term +E+E+E is a correction: Corollary 6 sums Lemma 5 over the epochs, and Lemma 5 needs M(Switches+1)M(\mathrm{Switches}+1)M(Switches+1) (see its note), with M=1M=1M=1 for losses in [0,1][0,1][0,1]. The general-perturbation form of Corollary 6 is not stated. Epochs are 0-based, so the sum runs over e=0,…,E−1e=0,\dots,E-1e=0,…,E−1.

Preamble
import Mathlib
import Definitions.Def_LimitedSwitching_HighProb_Setting

open MeasureTheory ProbabilityTheory
open scoped ENNReal
Formal statement
namespace LimitedSwitching.HighProb

theorem corollary_6 {n : ℕ} [NeZero n] (ω : ℕ → Fin n → ℝ) (hω : ∀ e i, 0 ≤ ω e i)
    (ε : ℝ) (hε : 0 < ε) (S' : ℝ) (T : ℕ) (hT : 1 ≤ T)
    (ℓ : ℕ → Fin n → ℝ) (hℓ : IsLossSeq ℓ) :
    regret T ℓ (play (batch (fplStar ε) S') ω ℓ)
      ≤ (switches T (play (batch (fplStar ε) S') ω ℓ) : ℝ)
        + (numEpochs (fplStar ε) S' ω ℓ T : ℝ)
        + ∑ e ∈ Finset.range (numEpochs (fplStar ε) S' ω ℓ T),
            Finset.univ.sup' Finset.univ_nonempty (ω e) / ε := by sorry

end LimitedSwitching.HighProb
Source
Altschuler & Talwar, Online learning over a finite action set with limited switching, arXiv:1803.01548v2, p. 13, Corollary 6, as specialised in the proof of Theorem 8, p. 14, display
Human review
  • Endorsed by Shuze Chen · Oct 19, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 19, 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