Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 10 — the representative (x(f), y(f)) of a pure stationary policy is an extreme point of the dual LP

Proved
HordijkKallenbergLP.ExtremePoint.theorem_10

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

extreme-pointlinear-programmingmarkov-decision-processmultichainp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1

Consider a Markov decision chain with finite state space EEE, finite nonempty action sets A(i)A(i)A(i) and transition probabilities piajp_{iaj}piaj​, and weights βj>0\beta_j>0βj​>0 with ∑jβj=1\sum_j\beta_j=1∑j​βj​=1. Let f∞f^\inftyf∞ be a pure stationary policy, f(i)∈A(i)f(i)\in A(i)f(i)∈A(i), and let (x(f),y(f))(x(f),y(f))(x(f),y(f)) be its representative (6):

xia(f)=[βTP∗(f)]i δaf(i),yia(f)=[βTD(f)+γTP∗(f)]i δaf(i).x_{ia}(f)=[\beta^TP^*(f)]_i\,\delta_{a f(i)},\qquad y_{ia}(f)=[\beta^TD(f)+\gamma^TP^*(f)]_i\,\delta_{af(i)}.xia​(f)=[βTP∗(f)]i​δaf(i)​,yia​(f)=[βTD(f)+γTP∗(f)]i​δaf(i)​.

Theorem 10. The representative (x(f),y(f))(x(f),y(f))(x(f),y(f)) is an extreme point of the feasible set of the dual linear program (3)–(5):

(x(f),y(f))∈ext⁡{(x,y) : (3),(4),(5)}.(x(f),y(f))\in\operatorname{ext}\{(x,y)\ :\ (3),(4),(5)\}.(x(f),y(f))∈ext{(x,y) : (3),(4),(5)}.

Under the correspondence of §3.3 between stationary policies and representatives, Theorem 10 places every pure stationary policy at a vertex of the dual feasible polyhedron. The paper remarks (p. 362, with an example in its reference [9]) that the converse fails: an extreme point may induce a non-pure policy.

Formalization Note. Being an extreme point includes feasibility. The variables are indexed by admissible pairs only. The vector γ\gammaγ uses the denominator ∑k∈Ejpki∗\sum_{k\in E_j}p^*_{ki}∑k∈Ej​​pki∗​ in place of the printed ∑kpki∗\sum_k p^*_{ki}∑k​pki∗​, a misprint under which the representative need not be feasible.

Preamble
import Mathlib
import Definitions.Def_MarkovDecisionProcesses_AverageReward
import Definitions.Def_BlackwellDiscreteDP_NearOne_LimitMatrix
import Definitions.Def_HordijkKallenbergLP_ExtremePoint_Model
open Matrix MarkovDecisionProcesses BlackwellDiscreteDP.NearOne
Formal statement
namespace HordijkKallenbergLP.ExtremePoint

/-- **Theorem 10.** Let `f^∞` be a pure and stationary policy. Then the corresponding
representative `(x(f), y(f))` is an extreme point of the dual linear programming problem.

Here `E` is finite, every `A(i)` is finite and nonempty, `β_j > 0` with `∑_j β_j = 1` (§2.1,
p. 353), the dual feasible set is (3)–(5) of p. 357, and `(x(f), y(f))` is (6) of p. 359 for the
rule `π_ia = 1` if `a = f(i)`, `0` otherwise.

Hordijk and Kallenberg, Linear Programming and Markov Decision Chains, Management Sci.
25(4):352–362 (1979), DOI 10.1287/mnsc.25.4.352, p. 361, Theorem 10.

**Formalization Note.** Being an extreme point (`Set.extremePoints ℝ`) includes being feasible.
The LP variables are indexed by admissible pairs only (`Pair M`), so non-admissible coordinates
are absent rather than free. `γ` in (6) uses the denominator `∑_{k ∈ E_j} p*_ki` in place of the
printed `∑_k p*_ki` (a misprint: with it the representative need not be feasible; see `gamma`). -/
theorem theorem_10 {S A : Type*} [Fintype S] [Fintype A] [DecidableEq S] [DecidableEq A]
    (M : StationaryMDP S A) (β : S → ℝ) (hβpos : ∀ j, 0 < β j) (hβsum : ∑ j, β j = 1)
    (f : S → A) (hf : ∀ i, f i ∈ M.admissible i) :
    (repX M β (pureRule f), repY M β (pureRule f)) ∈
      Set.extremePoints ℝ (dualFeasible M β) := by sorry

end HordijkKallenbergLP.ExtremePoint
Source
Hordijk and Kallenberg, Linear Programming and Markov Decision Chains, Management Sci. 25(4):352–362 (1979), DOI 10.1287/mnsc.25.4.352, p. 361, Theorem 10
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