Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 5.3, p. 22 — NPG with softmax policies from θ⁽⁰⁾ = 0: V⁽ᵀ⁾(ρ) ≥ V*(ρ) − log|A|/(ηT) − 1/((1−γ)²T)

Proved
PolicyGradTheory.NPG.npg_global_convergence

by mikedeng1 · Oct 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

global-convergencenatural-policy-gradientp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1softmax-policy

This is Theorem 5.3 (global convergence for natural policy gradient).

Let (P,r,γ)(P,r,\gamma)(P,r,γ) be a finite MDP with state set S\mathcal SS, nonempty action set A\mathcal AA, rewards in [0,1][0,1][0,1] and γ∈[0,1)\gamma\in[0,1)γ∈[0,1). Let ρ∈Δ(S)\rho\in\Delta(\mathcal S)ρ∈Δ(S) be a start distribution from which every state is reachable, and let π⋆\pi^\starπ⋆ be an optimal policy, with V⋆=Vπ⋆V^\star=V^{\pi^\star}V⋆=Vπ⋆. Fix a step size η>0\eta>0η>0, and run the natural policy gradient updates (15) with softmax policies, using ρ\rhoρ, from θ(0)=0\theta^{(0)}=0θ(0)=0:

θ(t+1)=θ(t)+η Fρ(θ(t))† ∇θV(t)(ρ).\theta^{(t+1)}=\theta^{(t)}+\eta\,F_\rho(\theta^{(t)})^{\dagger}\,\nabla_\theta V^{(t)}(\rho).θ(t+1)=θ(t)+ηFρ​(θ(t))†∇θ​V(t)(ρ).

Then for every T>0T>0T>0,

V(T)(ρ)  ≥  V⋆(ρ)−log⁡∣A∣ηT−1(1−γ)2T.V^{(T)}(\rho)\;\ge\;V^\star(\rho)-\frac{\log|\mathcal A|}{\eta T}-\frac{1}{(1-\gamma)^2T}.V(T)(ρ)≥V⋆(ρ)−ηTlog∣A∣​−(1−γ)2T1​.

The bound does not depend on the number of states or on any distribution mismatch coefficient, although θ↦Vπθ(ρ)\theta\mapsto V^{\pi_\theta}(\rho)θ↦Vπθ​(ρ) is not concave. It is the paper's headline rate for exact natural policy gradient.

Formalization Note V(T)V^{(T)}V(T) is the value of the softmax policy after TTT updates. log⁡\loglog is the natural logarithm. π⋆\pi^\starπ⋆ is given as a policy satisfying the published IsOptimalPolicy (its value dominates every policy's at every state). Two standing assumptions of the page are explicit binders: A≠∅\mathcal A\ne\emptysetA=∅ (implicit on the page, since Δ(A)\Delta(\mathcal A)Δ(A) is nonempty), and reachability of every state from ρ\rhoρ (§5.3, footnote 6). The Moore–Penrose pseudoinverse is the published predicate VectorSpaceOpt.IsPseudoinverse.

Preamble
import Mathlib
import Definitions.Def_PolicyGradTheory_ProjGA_MDP
import Definitions.Def_PolicyGradTheory_NPG_Algorithm
import Definitions.Def_FoundationsML_ReinforcementLearning_IsOptimalPolicy

open FoundationsML.ReinforcementLearning
Formal statement
namespace PolicyGradTheory.NPG

/-- Theorem 5.3 (global convergence for NPG; arXiv:1908.00261v5, p. 22): running the NPG
updates (15) with `ρ ∈ ∆(S)`, `θ^{(0)} = 0` and a fixed `η > 0`, for all `T > 0`,
`V^{(T)}(ρ) ≥ V^⋆(ρ) − log|A|/(ηT) − 1/((1−γ)²T)`. -/
theorem npg_global_convergence {S A : Type*} [Fintype S] [DecidableEq S] [Fintype A] [DecidableEq A]
    (P : S → A → S → ℝ) (r : S → A → ℝ) (γ : ℝ) (hM : PolicyGradTheory.ProjGA.IsFiniteMDP P r γ)
    [Nonempty A] (ρ : S → ℝ) (hρ : PolicyGradTheory.ProjGA.IsDist ρ) (hreach : AllReachable P γ ρ)
    (η : ℝ) (hη : 0 < η) (θ : ℕ → EuclideanSpace ℝ (S × A)) (hθ0 : θ 0 = 0)
    (hrun : IsNPGRun P r γ ρ η θ)
    (πstar : S → A → ℝ) (hπstar : IsPolicy πstar) (hopt : IsOptimalPolicy πstar P r γ) :
    ∀ T : ℕ, 0 < T →
      PolicyGradTheory.ProjGA.valueAt πstar P r γ ρ - Real.log (Fintype.card A : ℝ) / (η * T) -
          1 / ((1 - γ) ^ 2 * T) ≤
        PolicyGradTheory.ProjGA.valueAt (softmaxPolicy (θ T)) P r γ ρ := by sorry

end PolicyGradTheory.NPG
Source
arXiv:1908.00261v5, Theorem 5.3, p. 22
Human review
  • Endorsed by Shuze Chen · Oct 6, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 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