Theorem 4 — Bounded loss with k segments
ProvedFBMSP.theorem4_bounded_lossFor any nonnegative integrable prediction law and independent integrable mean-zero error, assume finite bounds almost surely. Then for every ,
No MHR assumption or nonnegativity of realized valuations is imposed. Prediction laws may be atomic or continuous. Taking the tight support bounds recovers the source statement; any enclosing almost-sure bounds give the same valid form.
import Definitions.Def_FBMSP_Model open MeasureTheory Set
namespace FBMSP
theorem theorem4_bounded_loss (ν η : Measure ℝ)
(hν : ValidPredictions ν) (hη : ValidNoise η)
(L U : ℝ) (hLU : L ≤ U) (hRange : ∀ᵐ x ∂ν, x ∈ Icc L U)
(k : ℕ) (hk : 1 ≤ k) :
segmentationLoss ν η k ≤ (U - L) / (k : ℝ) := by sorry
end FBMSPRead-back
What the Lean code literally says, in plain math · gpt-6-astra
Let be probability measures on for which the identity function is integrable, with for -almost every and . Write , , , and . Empty menus are included and yield zero; these quantities optimize over all nonnegative real prices and do not assume any maximizing price or menu has been chosen. For all real with and for -almost every , and every natural , the assertion is . There is no condition , although the assumptions require the interval to carry a probability law supported on nonnegative values and so cannot hold if . The case is included and gives the upper bound ; is excluded. No absolute-continuity, log-concavity, bounded-shock, or nonnegative-realized-valuation condition is imposed on .
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.