Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

(4.6)–(4.8), proof of Theorem 4.1, p. 18 — the Fenchel coupling to a dominated strategy grows at least linearly

Proved
ReinfRegGames.Extinction.eq_4_8

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

dominated-strategiesfenchel-couplinglearning-in-gamesp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1

Consider a finite game, a penalty function hℓh_\ellhℓ​ for every player ℓ\ellℓ, and an orbit x(t)=Q(y(t))x(t)=Q(y(t))x(t)=Q(y(t)) of the reinforcement learning dynamics (RL). Fix a player kkk and two mixed strategies pk,pk′∈Xkp_k,p'_k\in\mathcal X_kpk​,pk′​∈Xk​, and let δ\deltaδ be a real number such that

uk(pk′;z−k)−uk(pk;z−k) ≥ δfor every mixed profile z∈X.u_k(p'_k;z_{-k})-u_k(p_k;z_{-k})\ \ge\ \delta\qquad\text{for every mixed profile } z\in\mathcal X .uk​(pk′​;z−k​)−uk​(pk​;z−k​) ≥ δfor every mixed profile z∈X.

Let Vk(yk)=hk(pk)−hk(pk′)−⟨yk∣pk−pk′⟩V_k(y_k)=h_k(p_k)-h_k(p'_k)-\langle y_k|p_k-p'_k\rangleVk​(yk​)=hk​(pk​)−hk​(pk′​)−⟨yk​∣pk​−pk′​⟩ be the cross-coupling (4.6). Then for every t≥0t\ge0t≥0,

Fk(pk,yk(t)) ≥ Vk(yk(0))+δ t.F_k(p_k,y_k(t))\ \ge\ V_k(y_k(0))+\delta\,t .Fk​(pk​,yk​(t)) ≥ Vk​(yk​(0))+δt.

With δ=δk=min⁡x∈X{uk(pk′;x−k)−uk(pk;x−k)}\delta=\delta_k=\min_{x\in\mathcal X}\{u_k(p'_k;x_{-k})-u_k(p_k;x_{-k})\}δ=δk​=minx∈X​{uk​(pk′​;x−k​)−uk​(pk​;x−k​)} this is (4.8). When pkp_kpk​ is strictly dominated by pk′p'_kpk′​, δk>0\delta_k>0δk​>0 and the Fenchel coupling between pkp_kpk​ and the score of player kkk diverges to +∞+\infty+∞, which by Proposition C.4 drives play away from pkp_kpk​.

Formalization Note The minimum δk\delta_kδk​ is the largest admissible δ\deltaδ, so stating the bound for every admissible δ\deltaδ is equivalent to stating it for δk\delta_kδk​, without a real infimum. Here uk(p;z−k)u_k(p;z_{-k})uk​(p;z−k​) is the expected payoff of player kkk when zkz_kzk​ is replaced by ppp. The orbit is an orbit of (RL) in the differential form of the Model definition.

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

theorem eq_4_8 {ι : Type*} [Fintype ι] [DecidableEq ι] {A : ι → Type*}
    [∀ k, Fintype (A k)] [∀ k, DecidableEq (A k)]
    (u : ι → (∀ k, A k) → ℝ) (h : ∀ k, (A k → ℝ) → ℝ) (K : ι → ℝ)
    (hpen : ∀ k, IsPenalty (h k) (K k))
    (y x : ℝ → ∀ k, A k → ℝ) (horb : IsRLOrbit u h y x)
    (k : ι) (p p' : A k → ℝ) (hp : AGT.IsLottery p) (hp' : AGT.IsLottery p') (δ : ℝ)
    (hδ : ∀ z : ∀ ℓ, A ℓ → ℝ, AGT.IsMixedProfile z →
      δ ≤ AGT.expectedPayoff u (Function.update z k p') k -
        AGT.expectedPayoff u (Function.update z k p) k) :
    ∀ t : ℝ, 0 ≤ t →
      (h k p - h k p' - y 0 k ⬝ᵥ (p - p')) + δ * t ≤ fenchelCoupling (h k) p (y t k) := by sorry

end ReinfRegGames.Extinction
Source
Mertikopoulos & Sandholm, Learning in games via reinforcement and regularization, arXiv:1407.6267v2, p. 18, proof of Theorem 4.1, (4.6)–(4.8)
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