Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Proposition 7.50 — analytically measurable ε-optimal and universally measurable exact selectors for lower semianalytic functions

Proved
BertsekasShreve.AnalyticSelection.lower_semianalytic_selection

by mikedeng1 · Oct 5, 2026 · Mathlib 0df444a (Lean v4.33.1)

dynamic-programminglower-semianalyticmeasurable-selectionp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-bookp2o-v1

Let XXX and YYY be Borel spaces, D⊆X×YD\subseteq X\times YD⊆X×Y an analytic set, and f:D→R∗f:D\to R^*f:D→R∗ a lower semianalytic function. Define f∗:projX(D)→R∗f^*:\mathrm{proj}_X(D)\to R^*f∗:projX​(D)→R∗ by f∗(x)=inf⁡y∈Dxf(x,y)f^*(x)=\inf_{y\in D_x}f(x,y)f∗(x)=infy∈Dx​​f(x,y).

(a) For every ε>0\varepsilon>0ε>0 there exists an analytically measurable function φ:projX(D)→Y\varphi:\mathrm{proj}_X(D)\to Yφ:projX​(D)→Y such that Gr(φ)⊆D\mathrm{Gr}(\varphi)\subseteq DGr(φ)⊆D and, for all x∈projX(D)x\in\mathrm{proj}_X(D)x∈projX​(D),

f[x,φ(x)]≤{f∗(x)+εif f∗(x)>−∞,−1/εif f∗(x)=−∞.f[x,\varphi(x)]\le\begin{cases}f^*(x)+\varepsilon&\text{if }f^*(x)>-\infty,\\-1/\varepsilon&\text{if }f^*(x)=-\infty.\end{cases}f[x,φ(x)]≤{f∗(x)+ε−1/ε​if f∗(x)>−∞,if f∗(x)=−∞.​

(b) The set

I={x∈projX(D)∣f(x,yx)=f∗(x) for some yx∈Dx}I=\{x\in\mathrm{proj}_X(D)\mid f(x,y_x)=f^*(x)\text{ for some }y_x\in D_x\}I={x∈projX​(D)∣f(x,yx​)=f∗(x) for some yx​∈Dx​}

is universally measurable, and for every ε>0\varepsilon>0ε>0 there exists a universally measurable function φ:projX(D)→Y\varphi:\mathrm{proj}_X(D)\to Yφ:projX​(D)→Y such that Gr(φ)⊆D\mathrm{Gr}(\varphi)\subseteq DGr(φ)⊆D and, for all x∈projX(D)x\in\mathrm{proj}_X(D)x∈projX​(D),

f[x,φ(x)]=f∗(x)if x∈I,f[x,φ(x)]≤{f∗(x)+εif x∉I, f∗(x)>−∞,−1/εif x∉I, f∗(x)=−∞.f[x,\varphi(x)]=f^*(x)\quad\text{if }x\in I,\qquad f[x,\varphi(x)]\le\begin{cases}f^*(x)+\varepsilon&\text{if }x\notin I,\ f^*(x)>-\infty,\\-1/\varepsilon&\text{if }x\notin I,\ f^*(x)=-\infty.\end{cases}f[x,φ(x)]=f∗(x)if x∈I,f[x,φ(x)]≤{f∗(x)+ε−1/ε​if x∈/I, f∗(x)>−∞,if x∈/I, f∗(x)=−∞.​

This is the selection theorem used throughout Chapters 8–9 to construct ε-optimal and optimal policies in Borel-space dynamic programming without semicontinuity or compactness assumptions.

Formalization Note R∗R^*R∗ is EReal, and f∗(x)+εf^*(x)+\varepsilonf∗(x)+ε is EReal addition of a real ε\varepsilonε, which is unambiguous (it equals +∞+\infty+∞ when f∗(x)=+∞f^*(x)=+\inftyf∗(x)=+∞). The case split keeps the book's −1/ε-1/\varepsilon−1/ε bound where f∗(x)=−∞f^*(x)=-\inftyf∗(x)=−∞; "analytically measurable" is with respect to the σ-algebra AX\mathscr A_XAX​ generated by the analytic sets, not the power set.

Preamble
import Mathlib
import Definitions.Def_BertsekasShreve_AnalyticSelection_BorelSpace
import Definitions.Def_BertsekasShreve_AnalyticSelection_Measurability
import Definitions.Def_BertsekasShreve_AnalyticSelection_LowerSemianalytic

open MeasureTheory
Formal statement
namespace BertsekasShreve.AnalyticSelection

/-- **Proposition 7.50** (pp. 184–185). Let `X`, `Y` be Borel spaces, `D ⊆ X × Y` analytic and
`f : D → R*` lower semianalytic, with `f*(x) = inf_{y ∈ D_x} f(x, y)` on `proj_X(D)`.
(a) For every `ε > 0` there is an analytically measurable `φ : proj_X(D) → Y` with graph in `D`
and `f(x, φ x) ≤ f*(x) + ε` where `f*(x) > -∞`, `f(x, φ x) ≤ -1/ε` where `f*(x) = -∞`.
(b) The set `I` where the infimum is attained is universally measurable, and for every `ε > 0`
there is a universally measurable `φ : proj_X(D) → Y` with graph in `D`, `f(x, φ x) = f*(x)` on
`I`, and the bounds of (a) off `I`. -/
theorem lower_semianalytic_selection {X Y : Type*}
    [TopologicalSpace X] [MeasurableSpace X] [BorelSpace X] [IsBorelSpace X]
    [TopologicalSpace Y] [MeasurableSpace Y] [BorelSpace Y] [IsBorelSpace Y]
    (D : Set (X × Y)) (hD : AnalyticSet D) (f : D → EReal) (hf : IsLowerSemianalytic D f) :
    -- (a)
    (∀ ε : ℝ, 0 < ε →
      ∃ (φ : (Prod.fst '' D) → Y) (hφ : ∀ x : Prod.fst '' D, ((x : X), φ x) ∈ D),
        IsAnalyticallyMeasurableFun (Prod.fst '' D) φ ∧
        ∀ x : Prod.fst '' D,
          (partialInf D f x ≠ ⊥ → f ⟨((x : X), φ x), hφ x⟩ ≤ partialInf D f x + (ε : EReal)) ∧
          (partialInf D f x = ⊥ → f ⟨((x : X), φ x), hφ x⟩ ≤ ((-1 / ε : ℝ) : EReal))) ∧
    -- (b)
    (IsUniversallyMeasurable (attainSet D f) ∧
      ∀ ε : ℝ, 0 < ε →
        ∃ (φ : (Prod.fst '' D) → Y) (hφ : ∀ x : Prod.fst '' D, ((x : X), φ x) ∈ D),
          IsUniversallyMeasurableFun (Prod.fst '' D) φ ∧
          ∀ x : Prod.fst '' D,
            ((x : X) ∈ attainSet D f → f ⟨((x : X), φ x), hφ x⟩ = partialInf D f x) ∧
            ((x : X) ∉ attainSet D f → partialInf D f x ≠ ⊥ →
              f ⟨((x : X), φ x), hφ x⟩ ≤ partialInf D f x + (ε : EReal)) ∧
            ((x : X) ∉ attainSet D f → partialInf D f x = ⊥ →
              f ⟨((x : X), φ x), hφ x⟩ ≤ ((-1 / ε : ℝ) : EReal))) := by sorry

end BertsekasShreve.AnalyticSelection
Source
Bertsekas & Shreve, Stochastic Optimal Control: The Discrete-Time Case, Athena Scientific 1996, pp. 184–185, Proposition 7.50 (Eqs. (114)–(116) of Chapter 7)
Human review
  • Endorsed by Shuze Chen · Oct 5, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 5, 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