Corollary 33 — online rounding with exact marginals
ProvedPagingDelay.PMD.corollary_33delayonline-algorithmsp2o-batch-stoc27ap2o-gran-per-chapterp2o-plan-paperp2o-v1paging
For dyadic page weights and an initial cache of exactly pages, there is one randomized rounding rule defined on all valid fractional cache histories, chosen before the history is supplied. Write . The rule depends only on this prefix and its seed; every membership event is measurable. It satisfies and , and
This is the rounding theorem to be proved; no BBN07 assumption is made.
Differences from the source:
- encoding: pages indexed by 1 through n → pages indexed by Fin n; reason: a bijective reindexing of the same finite page universe
- encoding: online random cache choices for vectors arriving one at a time → one probability space and a measurable prefix map R, selected before every finite history; reason: the full random tape is the seed; a prefix map cannot depend on future vectors, and its jointly defined outputs retain transition correlations
Preamble
import Definitions.Def_PagingDelay_PMD_Rounding
Formal statement
namespace PagingDelay.PMD
open scoped BigOperators
/-- Julian Born, Yann Disser, Linda Thelen, Tight weighted paging with delay via mirror descent, STOC 2027 submission #453 (October 2026), Corollary 33 — online rounding with exact marginals, p. 26. -/
theorem corollary_33 (n k : ℕ) (hk : 1 ≤ k) (hkn : k ≤ n)
(w : Fin n → ℝ) (hw : DyadicWeights w)
(O₀ : Finset (Fin n)) (hO₀ : O₀.card = k) :
∃ R : OnlineRounding n,
(∀ ω, R.choose ω [indicator O₀] = O₀) ∧
(∀ ys, CacheHistory k O₀ ys →
(∀ ω, (R.choose ω ys).card = k) ∧
(∀ i, roundingMarginal R ys i = ENNReal.ofReal (ys.getLastD (indicator O₀) i)) ∧
(∀ y, FractionalCache k y → roundingFetch w R ys y ≤
ENNReal.ofReal (5 * ∑ i, w i * |y i - ys.getLastD (indicator O₀) i|))) := by sorry
end PagingDelay.PMD
Source
Julian Born, Yann Disser, Linda Thelen, Tight weighted paging with delay via mirror descent, STOC 2027 submission #453 (October 2026), Corollary 33 — online rounding with exact marginals, p. 26. https://github.com/sigact-stoc/stoc2027/releases/download/papers-1/453.pdf
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.