Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma 4 — for g ∈ E(f), y(g) = y(f) + w(g) with w(g) the unique solution of (10)

Proved
VeinottNoDiscount.Improve.lemma_4

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

bias-optimalitymarkov-decision-processp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1policy-improvement

Let f∈Ff\in Ff∈F and g∈E(f)g\in E(f)g∈E(f), the set of decision rules for which the gain and bias tests (i), (ii) hold with equality at every state. Then the system

[I−Q(g)] w=0,Q∗(g) w=Q∗(g) (−y(f))(10)[I-Q(g)]\,w=0,\qquad Q^*(g)\,w=Q^*(g)\,(-y(f))\tag{10}[I−Q(g)]w=0,Q∗(g)w=Q∗(g)(−y(f))(10)

has exactly one solution w(g)w(g)w(g), and

y(g)=y(f)+w(g).(11)y(g)=y(f)+w(g).\tag{11}y(g)=y(f)+w(g).(11)

The lemma turns the maximization of the bias over E(f)E(f)E(f) into the maximization of a gain-type quantity, w(g)w(g)w(g), which has the same form as the original gain maximization with reward −y(f)-y(f)−y(f).

Formalization Note w(g)w(g)w(g) is defined as Q∗(g)(−y(f))Q^*(g)(-y(f))Q∗(g)(−y(f)); the statement asserts that this vector is the one and only solution of (10).

Preamble
import Mathlib
import Definitions.Def_BlackwellDiscreteDP_NearOne_LimitMatrix
import Definitions.Def_BlackwellDiscreteDP_NearOne_Model
import Definitions.Def_BlackwellDiscreteDP_NearOne_GainBias
import Definitions.Def_VeinottNoDiscount_Improve_Sets
open Filter Topology Matrix BlackwellDiscreteDP.NearOne
Formal statement
namespace VeinottNoDiscount.Improve

/-- **Lemma 4.** Suppose `f ε F` and `g ε E(f)`. Let `w(g)` be the unique solution to
(10) `[I − Q(g)]w = 0, Q*(g)w = Q*(g)(−y(f))`. Then (11) `y(g) = y(f) + w(g)`.

Veinott, On Finding Optimal Policies in Discrete Dynamic Programming with No Discounting,
Ann. Math. Statist. 37(5):1284–1294 (1966), DOI 10.1214/aoms/1177699272, p. 1289, Lemma 4, (10), (11).

**Formalization Note.** `w M f g` is the closed form `Q*(g)(−y(f))`; the first conjunct says it is
the one and only solution of (10) (the paper's "unique solution"), the second is (11). -/
theorem lemma_4 {St Act : Type*} [Fintype St] [DecidableEq St] [Nonempty St]
    [Fintype Act] [Nonempty Act] (M : Model St Act)
    (f g : St → Act) (hg : g ∈ ESet M f) :
    (∀ v : St → ℝ,
        ((1 - M.Q g) *ᵥ v = 0 ∧ M.Qstar g *ᵥ v = M.Qstar g *ᵥ (-(M.y f))) ↔ v = w M f g) ∧
      M.y g = M.y f + w M f g := by sorry

end VeinottNoDiscount.Improve
Source
Veinott, On Finding Optimal Policies in Discrete Dynamic Programming with No Discounting, Ann. Math. Statist. 37(5):1284–1294 (1966), DOI 10.1214/aoms/1177699272, p. 1289, Lemma 4, (10), (11)
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