Proof of Theorem 1, p. 1332 — problem (13) is separable in γ (F-a.e.)
ProvedFragLOB.Collapse.proof_thm1_separableLet be a probability distribution of investor types on with finite mean . Fix arbitrary delays and write for the payoff of a type- investor using the routing vector . Let be a routing decision (measurable, with in the simplex for every ). Then maximizes the aggregate payoff over all routing decisions,
if and only if, for -almost every type , maximizes over the simplex .
This is the separability of problem (13) in invoked in the proof of Theorem 1: it turns optimality of the aggregated problem into individual rationality (10), type by type, and back.
Formalization Note. The pointwise statement holds only -almost everywhere: changing on an -null set of types does not change any integral, so optimality in (13) cannot force rationality at every single type (for , , , , and exponential, altering the threshold policy at keeps it optimal but makes type irrational). The finite mean makes every objective a genuine (integrable) integral; without it Lean's integral of a non-integrable function would be .
import Mathlib import Definitions.Def_FragLOB_Collapse_Model open MeasureTheory
namespace FragLOB.Collapse
theorem proof_thm1_separable {N : ℕ} (M : Market N) (F : Measure ℝ) [IsProbabilityMeasure F]
(hF : F (Set.Iio 0) = 0) (hmean : Integrable (fun γ => γ) F) (d : Fin N → ℝ)
(pi : ℝ → Fin (N + 1) → ℝ) (hpi : IsRouting pi) :
(∀ pi' : ℝ → Fin (N + 1) → ℝ, IsRouting pi' →
∫ γ, payoff M d γ (pi' γ) ∂F ≤ ∫ γ, payoff M d γ (pi γ) ∂F) ↔
∀ᵐ γ ∂F, IsBestResponse M d γ (pi γ) := by sorry
end FragLOB.Collapse
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.