Theorem 5.2(a) (N) — the operator is monotone
ProvedNegativeDP.OptEq.theorem52abellman-operatordynamic-programmingnegative-dynamic-programmingp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be a Markov policy of a negative dynamic programming problem, the operator of , and its associated operator on , the non-positive extended real-valued Borel functions on . For ,
Monotonicity of is used to propagate the conservation property along the iterates (Lemma 6.2).
The paper states Theorem 5.2 for the discounted, positive and negative cases; this item is the negative case. Formalization Note Inequalities are pointwise; and are explicit hypotheses (IsNegM).
Preamble
import Mathlib import Definitions.Def_DiscountedDP_Stationary_Model import Definitions.Def_DiscountedDP_Stationary_Return import Definitions.Def_DiscountedDP_Stationary_Operators import Definitions.Def_NegativeDP_OptEq_Model open MeasureTheory ProbabilityTheory Filter Topology open scoped ENNReal open DiscountedDP.Stationary (Hist Plan MarkovPlan stationary IsGenerated IsGeneratedPlan)
Formal statement
namespace NegativeDP.OptEq
variable {S A : Type*} [MeasurableSpace S] [StandardBorelSpace S] [Nonempty S]
[MeasurableSpace A] [StandardBorelSpace A] [Nonempty A]
/-- Strauch (1966), Theorem 5.2(a), p. 879, negative case: `U` is monotone on `M(S)`. -/
theorem theorem52a (P : NegativeDP.Stationary.Problem S A) (π : MarkovPlan S A) (u v : S → EReal)
(hu : NegativeDP.Stationary.IsNegM u) (hv : NegativeDP.Stationary.IsNegM v) (huv : ∀ s, u s ≤ v s) :
∀ s, U P π u s ≤ U P π v s := by sorry
end NegativeDP.OptEq
Source
Strauch, Negative Dynamic Programming, Ann. Math. Statist. 37 (1966), p. 879, Theorem 5.2(a)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.