Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 9.1 (N) — essentially finite actions: U^n0 → v* and a stationary optimal policy exists

Proved
NegativeDP.EssFinite.theorem91

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

dynamic-programmingnegative-dynamic-programmingp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1value-iteration

Consider Strauch's negative dynamic programming problem: non-empty Borel state and action spaces SSS, AAA, a law of motion q(⋅∣s,a)q(\cdot\mid s,a)q(⋅∣s,a), and a Borel return r(s,a,t)≤0r(s,a,t)\le0r(s,a,t)≤0 with ∫r(s,a,t) dq(t∣s,a)>−∞\int r(s,a,t)\,dq(t\mid s,a)>-\infty∫r(s,a,t)dq(t∣s,a)>−∞, without discounting. Suppose that AAA is essentially finite by the Markov policy π∗=(f1,f2,… )\pi^*=(f_1,f_2,\dots)π∗=(f1​,f2​,…): there is a partition of SSS into Borel sets S1,S2,…S_1,S_2,\dotsS1​,S2​,… such that for every (s,a)(s,a)(s,a) with s∈Sns\in S_ns∈Sn​ one of f1(s),…,fn(s)f_1(s),\dots,f_n(s)f1​(s),…,fn​(s) is equivalent to aaa at sss (same return function r(s,⋅,⋅)r(s,\cdot,\cdot)r(s,⋅,⋅) and same transition law). Let UUU be the operator associated with π∗\pi^*π∗, Uu=sup⁡nTnuUu=\sup_n T_nuUu=supn​Tn​u with Tnu(s)=∫[r(s,fn(s),t)+u(t)] dq(t∣s,fn(s))T_nu(s)=\int[r(s,f_n(s),t)+u(t)]\,dq(t\mid s,f_n(s))Tn​u(s)=∫[r(s,fn​(s),t)+u(t)]dq(t∣s,fn​(s)). Then

  1. value iteration from zero converges to the optimal return: for every state sss,
Un0(s)⟶v∗(s)=sup⁡πI(π)(s)(n→∞);U^n0(s)\longrightarrow v^*(s)=\sup_\pi I(\pi)(s)\qquad(n\to\infty);Un0(s)⟶v∗(s)=πsup​I(π)(s)(n→∞);
  1. there exists a stationary optimal policy f(∞)f^{(\infty)}f(∞), i.e. a measurable f:S→Af:S\to Af:S→A with I(f(∞))(s)≥sup⁡πI(π)(s)I(f^{(\infty)})(s)\ge\sup_\pi I(\pi)(s)I(f(∞))(s)≥supπ​I(π)(s) for every sss.

The paper states this for the discounted and negative cases; this item is the negative case. In particular, when AAA is finite there is an optimal policy (§1). Without such a hypothesis, Example 6.1 of the paper shows that in the negative case lim⁡nUn0\lim_n U^n0limn​Un0, the supremum of III over the policies generated by the rules of UUU, and the supremum over the corresponding stationary policies may all differ.

Formalization Note The convergence in (1) is pointwise in EReal, so the limit −∞-\infty−∞ is allowed; U00=0U^0 0=0U00=0. v∗v^*v∗ is the supremum over every randomized history-dependent plan, and optimality is against every plan at every state. Essential finiteness is zero-based: Lean's piece n is the paper's Sn+1S_{n+1}Sn+1​ and allows f1,…,fn+1f_1,\dots,f_{n+1}f1​,…,fn+1​; the pieces are measurable, pairwise disjoint and cover SSS, and may be empty.

Preamble
import Mathlib
import Definitions.Def_DiscountedDP_Stationary_Operators
import Definitions.Def_NegativeDP_EssFinite_Model

open MeasureTheory ProbabilityTheory Filter Topology
open DiscountedDP.Stationary (Hist Plan MarkovPlan MarkovPlan.toPlan stationary IsGenerated IsGeneratedPlan)
Formal statement
namespace NegativeDP.EssFinite

variable {S A : Type*} [MeasurableSpace S] [StandardBorelSpace S] [Nonempty S]
  [MeasurableSpace A] [StandardBorelSpace A] [Nonempty A]

/-- Theorem 9.1 (N), Strauch (1966), p. 887. If `A` is essentially finite by
`π* = (f_1, f_2, …)` and `U` is the operator associated with `π*`, then
(a) `U^n 0 → v* = sup_π NegativeDP.Stationary.I(π)` pointwise, and
(b) there exists a stationary optimal policy `f^(∞)`. -/
theorem theorem91 (P : NegativeDP.Stationary.Problem S A) (πstar : MarkovPlan S A)
    (h : EssentiallyFinite P πstar) :
    (∀ s, Tendsto (fun n => (NegativeDP.OptEq.U P πstar)^[n] 0 s) atTop (𝓝 (NegativeDP.Stationary.vstar P s))) ∧
      ∃ f : {g : S → A // Measurable g}, NegativeDP.Stationary.IsOptimal P (stationary f).toPlan := by sorry

end NegativeDP.EssFinite
Source
Strauch, Negative Dynamic Programming, Ann. Math. Statist. 37 (1966), p. 887, Theorem 9.1
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