Corollary 4.1 — deterministic optimal-value stability
ProvedInventoryBounds.minimizer_value_errorFor true and empirical minimizers over a common nonempty candidate class, a uniform objective error at most b gives an absolute difference at most b between their minimum values. The factor is one for values, compared with two for policy regret.
import Definitions.Def_InventoryBounds_Primitives import Mathlib.Tactic.Linarith import Mathlib.Tactic.Ring set_option autoImplicit false
namespace InventoryBounds
theorem minimizer_value_error {α : Type*} (f fhat : α → ℝ)
(s shat : α) (b : ℝ)
(hdev : ∀ a, |f a - fhat a| ≤ b)
(htrue : ∀ a, f s ≤ f a)
(hemp : ∀ a, fhat shat ≤ fhat a) :
|fhat shat - f s| ≤ b := by sorry
end InventoryBoundsRead-back
What the Lean code literally says, in plain math · gpt-6
For every type in any universe, every two functions , every two elements , and every real , suppose for all , for all , and for all . Then . Thus the assumptions supply a global minimizer of and a global minimizer of ; neither is required to be unique, and they may coincide. No structure or finiteness assumption on or explicit nonnegativity assumption on is present. The supplied elements rule out an empty , and the uniform absolute-error bound therefore forces , making the hypotheses impossible for .
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.