Answer of the strong oracle
DefinitionBinaryQueryLB_Supplement_strongAnswerconvex-optimizationlower-boundsoracle-complexityp2o-gran-per-chapterp2o-plan-paperp2o-v1
Under a first-order chart , the strong oracle answers the query on the function by
the exact value of at together with one exact linear measurement of the reported subgradient, with the Euclidean inner product of .
Definition code
import Mathlib
import Definitions.Def_BinaryQueryLB_Chart
namespace BinaryQueryLB.Supplement
/-- **Answer of the strong oracle O\*.** Shen–Zhong, *Near-Quadratic Lower Bounds for Convex Optimization with Binary First-Order Queries* (manuscript, sha256 9502d824…), §2.2, p.12: "The oracle `O∗` answers a query `(y, v)` with `y ∈ X` and `v ∈ R^d` by the exact pair
`(f(y), ⟨v, g_f(y)⟩)`", where `g_f(y) = G(f, y)` under the chart `G` (Definition 2.2, p.11).
`strongAnswer G f (y, v) = (f y, ⟨v, G f y⟩)`, with the Euclidean inner product of ℝ^d.
**Formalization Note.** ℝ^d is `EuclideanSpace ℝ (Fin d)`, whose inner product is the Euclidean
one `∑ᵢ vᵢ gᵢ`. -/
noncomputable def strongAnswer {d : ℕ} (G : BinaryQueryLB.Chart d)
(f : EuclideanSpace ℝ (Fin d) → ℝ)
(q : EuclideanSpace ℝ (Fin d) × EuclideanSpace ℝ (Fin d)) : ℝ × ℝ :=
(f q.1, inner ℝ q.2 (G f q.1))
end BinaryQueryLB.Supplement
Source
Shen–Zhong, Near-Quadratic Lower Bounds for Convex Optimization with Binary First-Order Queries (manuscript, sha256 9502d824…), §2.2, p.12