Measurability and integrability of policy terminal reward and cost
ProvedPandoraBO.Private20261023.policy_terminal_measurable_integrableexpected-budgetmeasure-theorypandoras-box
For every policy in the finite independent-box model, terminal reward and terminal opening cost are measurable and integrable under the joint law of all box rewards and the independent random seed. Policies are measurable and adaptive as specified by the model; no continuity or atomlessness assumption on reward distributions is added. This ensures the expectations used in the budgeted search problem are ordinary finite expectations.
Preamble
import Definitions.Def_PandoraBO_Model open MeasureTheory set_option autoImplicit false
Formal statement
theorem PandoraBO.Private20261023.policy_terminal_measurable_integrable {n : ℕ} (M : PandoraBO.Instance n) (π : PandoraBO.Policy n) :
Measurable (PandoraBO.terminalReward π) ∧
Measurable (PandoraBO.terminalCost M π) ∧
Integrable (PandoraBO.terminalReward π) (PandoraBO.jointLaw M) ∧
Integrable (PandoraBO.terminalCost M π) (PandoraBO.jointLaw M) := by sorry
Source
Probability foundations of the finite-box policy model underlying Xie et al., author revised manuscript reviewed 23 October 2026, Appendix B.5 (Theorem 2 proof). This auxiliary measurability/integrability result is isolated for formalization, not a separately numbered source lemma.