Integration under a rescaled uniform randomization seed
ProvedPandoraBO.Private20261023.uniform_seed_mixture_integralexpected-budgetmeasure-theorypandoras-box
Let μ be a sigma-finite-in-the-S-finite-sense measure, and let the seed be uniformly distributed on [0,1], independently of the first coordinate. Let 0<a<1 and let f, g, and F be integrable under the product law, where F uses f(x,u/a) when u≤a and g(x,(u−a)/(1−a)) otherwise. Then the integral of F is a times the integral of f plus (1−a) times the integral of g. This change-of-variables identity supplies the randomization step for policy mixtures.
Preamble
import Definitions.Def_PandoraBO_Model import Mathlib.MeasureTheory.Integral.Prod open MeasureTheory PandoraBO set_option autoImplicit false
Formal statement
theorem PandoraBO.Private20261023.uniform_seed_mixture_integral {X : Type*} [MeasurableSpace X]
(μ : Measure X) [SFinite μ] (a : ℝ) (ha0 : 0 < a) (ha1 : a < 1)
(f g F : X × ℝ → ℝ)
(hformula : ∀ x u, F (x, u) =
if u ≤ a then f (x, u / a) else g (x, (u - a) / (1 - a)))
(hf : Integrable f (μ.prod PandoraBO.seedLaw))
(hg : Integrable g (μ.prod PandoraBO.seedLaw))
(hF : Integrable F (μ.prod PandoraBO.seedLaw)) :
(∫ ω, F ω ∂μ.prod PandoraBO.seedLaw) =
a * (∫ ω, f ω ∂μ.prod PandoraBO.seedLaw) +
(1 - a) * (∫ ω, g ω ∂μ.prod PandoraBO.seedLaw) := by sorry
Source
Auxiliary construction for Xie et al., Cost-aware Bayesian Optimization via the Pandora’s Box Gittins Index, author revised manuscript reviewed 23 October 2026, Appendix B.5, Lemmas 7–8 and Theorem 2. This result isolates the measurable-policy existence or policy-randomization step; it is not a verbatim numbered claim.