Section 3.2, Eq. (4) — mean preservation: E[P | S] = S
ProvedCalibratedMA.Randomized.eq_4calibrationp2o-batch-stoc27ap2o-gran-per-chapterp2o-plan-paperp2o-v1randomized-rounding
Let be a grid partition. The mean-preserving rounding preserves the conditional mean: for every ,
Mean preservation is why rounding leaves the multiaccuracy residuals unchanged in expectation; only the calibration residual pays for rounding.
Formalization Note. The statement is pointwise in : the integral of the identity against the law roundMeasure M γ s equals .
Differences from the source:
- encoding: → for every ; reason: given , has law , so the pointwise identity is the conditional one.
- encoding: with an integer → , ; reason: same instances.
Preamble
import Mathlib import Definitions.Def_CalibratedMA_Randomized_Rounding open MeasureTheory
Formal statement
namespace CalibratedMA.Randomized
/-- Section 3.2, Eq. (4) (p. 13): mean-preserving rounding, `E[Round_Γ(s)] = s` for every
`s ∈ [0,1]`, i.e. `E[P | S] = S` for `P = Round_Γ(S)`. -/
theorem eq_4 (K M : ℕ) (γ : ℕ → ℝ) (hΓ : IsGridPartition K M γ) (s : ℝ)
(hs : s ∈ Set.Icc (0 : ℝ) 1) :
∫ p, p ∂(roundMeasure M γ s) = s := by sorry
end CalibratedMA.Randomized
Source
Jiuyao Lu, Georgy Noarov, Aaron Roth, Calibrated Multiaccuracy: Deterministic and Randomized Sample Complexity, STOC 2027 submission #900 (October 2026), p. 13, Section 3.2, Eq. (4)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.