Weighted Cauchy–Schwarz after grouping by rank
ProvedNarrowDNF.RankGrouping.weighted_squared_sum_le_rank_groupscombinatoricsnarrow-dnf-authors-importprobabilitystoc-2027
Let be finite with weights . Let be finite sets and a rank assignment; equivalently, every index in has its rank in . For arbitrary real values ,
There is no normalization assumption on . Taking the rank set to be gives the factor ; allowing rank zero also gives the factor .
Preamble
import Mathlib.Algebra.BigOperators.Ring.Finset
import Mathlib.Algebra.Order.BigOperators.Ring.Finset
import Mathlib.Data.Fintype.BigOperators
import Mathlib.Data.Fintype.Pi
import Mathlib.Data.Real.Basic
import Mathlib.Logic.Equiv.Prod
import Mathlib.Tactic.Linarith
import Mathlib.Tactic.Ring
import Definitions.Def_NarrowDNF_ProductMeasure
namespace NarrowDNF.ProductMeasure
end NarrowDNF.ProductMeasure
namespace NarrowDNF.RankGrouping
end NarrowDNF.RankGrouping
open NarrowDNF
open NarrowDNF.RankGrouping
open scoped BigOperators
variable {α β : Type*} [DecidableEq β]
Formal statement
theorem NarrowDNF.RankGrouping.weighted_squared_sum_le_rank_groups {Ω : Type*} [Fintype Ω]
(weights : Ω → ℝ) (hw : ∀ x, 0 ≤ weights x)
(A : Finset α) (R : Finset β) (rank : α → β)
(hcover : ∀ i ∈ A, rank i ∈ R) (v : Ω → α → ℝ) :
(∑ x, weights x * (∑ i ∈ A, v x i)^2) ≤ (R.card : ℝ) *
∑ r ∈ R, ∑ x, weights x * (∑ i ∈ A with rank i = r, v x i)^2 := by sorrySource