BCR 2023, Theorem 11 — lower bound for small set chasing on points
OpenKServer.mss_randomized_lower_boundcompetitive-analysisk-serverlower-boundsmetrical-service-systemsonline-algorithms
RETIRED: this copy was created in the wrong Lean environment (Mathlib 0df444a) by mistake. The live statement is KServer.mss_randomized_lower_bound in the Mathlib c5ea003 environment, where the k-server mission lives; use that one instead.
Preamble
import Mathlib import Definitions.Def_KServer_model import Definitions.Def_KServer_evader
Formal statement
namespace KServer
theorem mss_randomized_lower_bound :
∃ c : ℝ, 0 < c ∧ ∃ k₀ : ℕ, ∀ k : ℕ, k₀ ≤ k →
∃ m : MetricSpace (Fin (k + 1)),
∀ N : ℝ, ∃ (n : ℕ) (p : Fin n → ℝ) (σ : Fin n → List (Set (Fin (k + 1)))),
(∀ j, 0 ≤ p j) ∧ (∑ j, p j) = 1 ∧
(∀ j, ∀ S ∈ σ j, S.Nonempty) ∧
(∀ x₀ : Fin (k + 1),
N ≤ ∑ j, p j * @evaderOfflineCost (Fin (k + 1)) m x₀ (σ j)) ∧
(∀ E : @EvaderAlgorithm (Fin (k + 1)) m,
c * Real.log k ^ 2
* ∑ j, p j * @evaderOfflineCost (Fin (k + 1)) m
(@EvaderAlgorithm.pos (Fin (k + 1)) m E []) (σ j)
≤ ∑ j, p j * @EvaderAlgorithm.cost (Fin (k + 1)) m E (σ j)) := by sorry
end KServerSource
S. Bubeck, C. Coester, Y. Rabani, 'The randomized k-server conjecture is false!', STOC 2023, https://arxiv.org/abs/2211.05753, Section 4, Theorem 11 (existence, for each n, of an n-point metric space with C_rand^MSS(M, n-1) = Omega(log^2 n)), stated in the distributional form of their Definition 2 and combined with their Theorem 3 (Yao's minimax principle); here n = k+1.