Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Proposition C.3, p. 35 — Fh(p,y)≥12K∥Q(y)−p∥2F_h(p,y)\ge\frac12K\|Q(y)-p\|^2Fh​(p,y)≥21​K∥Q(y)−p∥2

Proved
ReinfRegGames.Extinction.proposition_C_3

by mikedeng1 · Oct 9, 2026 · Mathlib 0df444a (Lean v4.33.1)

convex-analysisfenchel-couplingp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1strong-convexity

Let hhh be a penalty function on the simplex Δ=Δ(B)\Delta=\Delta(B)Δ=Δ(B) with strong convexity constant K>0K>0K>0 (Definition 2.1, Euclidean norm), let p∈Δp\in\Deltap∈Δ, and let y∈RBy\in\mathbb R^By∈RB be a score vector with x=Q(y)x=Q(y)x=Q(y), i.e. x∈Δx\in\Deltax∈Δ maximizes ⟨y∣x′⟩−h(x′)\langle y|x'\rangle-h(x')⟨y∣x′⟩−h(x′) over Δ\DeltaΔ. Then the Fenchel coupling Fh(p,y)=h(p)+h∗(y)−⟨y∣p⟩F_h(p,y)=h(p)+h^*(y)-\langle y|p\rangleFh​(p,y)=h(p)+h∗(y)−⟨y∣p⟩ satisfies

Fh(p,y) ≥ 12K ∥Q(y)−p∥22.F_h(p,y)\ \ge\ \tfrac12K\,\|Q(y)-p\|_2^2 .Fh​(p,y) ≥ 21​K∥Q(y)−p∥22​.

In particular Fh(p,y)≥0F_h(p,y)\ge0Fh​(p,y)≥0, the form in which the inequality enters the proof of Theorem 4.1 (as Fk(pk′,yk)≥0F_k(p'_k,y_k)\ge0Fk​(pk′​,yk​)≥0, before (4.8)). The bound holds at every yyy, including those for which Q(y)Q(y)Q(y) lies on the boundary of the simplex outside Δp\Delta_pΔp​. Moreover, for every sequence of score vectors yjy_jyj​ with xj=Q(yj)x_j=Q(y_j)xj​=Q(yj​),

Fh(p,yj)→0  ⟺  xj→p.F_h(p,y_j)\to0 \iff x_j\to p .Fh​(p,yj​)→0⟺xj​→p.

Formalization Note This item is the first sentence of Proposition C.3 (the inequality and the "Fh(p,y)→0F_h(p,y)\to0Fh​(p,y)→0 if and only if Q(y)→pQ(y)\to pQ(y)→p" clause, read along sequences); (C.11) is the separate item eq_C_11. The norm is the Euclidean norm of the definition of penalty function.

Preamble
import Mathlib
import Definitions.Def_agt_games
import Definitions.Def_ReinfRegGames_Extinction_Model
Formal statement
namespace ReinfRegGames.Extinction

theorem proposition_C_3 {B : Type*} [Fintype B] (h : (B → ℝ) → ℝ) (K : ℝ) (hpen : IsPenalty h K)
    (p : B → ℝ) (hp : p ∈ stdSimplex ℝ B) (y x : B → ℝ) (hQ : IsChoice h y x) :
    1 / 2 * K * sqDist x p ≤ fenchelCoupling h p y ∧
      ∀ ys xs : ℕ → B → ℝ, (∀ j, IsChoice h (ys j) (xs j)) →
        (Filter.Tendsto (fun j => fenchelCoupling h p (ys j)) Filter.atTop (nhds 0) ↔
          Filter.Tendsto xs Filter.atTop (nhds p)) := by sorry

end ReinfRegGames.Extinction
Source
Mertikopoulos & Sandholm, Learning in games via reinforcement and regularization, arXiv:1407.6267v2, p. 35, Proposition C.3 (first sentence; proof (C.12)–(C.14))
Human review
  • Endorsed by Shuze Chen · Oct 9, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 9, 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