import Distributed2Coloring.Definitions import Distributed2Coloring.Reduction /-! # Distributed2Coloring: Main results This file collects the public-facing theorems connecting: - the certified finite lower bound at `n = 1_000_000`, and - one-round `ClassicalAlgorithm`s on the directed cycle. We also package these bounds as statements about an infimum `p⋆` over all measurable local rules. -/ namespace Distributed2Coloring open scoped unitInterval namespace ClassicalAlgorithm /-- The optimal monochromatic-edge probability over all one-round `ClassicalAlgorithm`s. This corresponds to the quantity `p⋆ := inf_f p(f)` in `theory/report/manuscript.tex`, where the infimum ranges over all measurable local rules. -/ noncomputable def pStar : ENNReal := sInf (Set.range ClassicalAlgorithm.p) /-- If `c` is a lower bound for `ClassicalAlgorithm.p` for *every* algorithm, then `c ≤ pStar`. -/ theorem le_pStar_of_forall_le {c : ENNReal} (hc : ∀ alg : ClassicalAlgorithm, c ≤ alg.p) : c ≤ ClassicalAlgorithm.pStar := by -- `c ≤ sInf (range p)` since `c` is a lower bound for every element of the range. refine le_sInf ?_ rintro _ ⟨alg, rfl⟩ exact hc alg /-- If there exists an algorithm achieving `p ≤ c`, then `pStar ≤ c`. In particular, any explicit construction gives an upper bound on `pStar`. -/ theorem pStar_le_of_exists_le {c : ENNReal} (hc : ∃ alg : ClassicalAlgorithm, alg.p ≤ c) : ClassicalAlgorithm.pStar ≤ c := by rcases hc with ⟨alg, halg⟩ have hpStar_le : ClassicalAlgorithm.pStar ≤ alg.p := by exact sInf_le ⟨alg, rfl⟩ exact hpStar_le.trans halg end ClassicalAlgorithm /-! ## Numerical bounds (as in the manuscript) The bridge step is `Distributed2Coloring.p_ge_23879` from `Distributed2Coloring/Reduction.lean`. -/ /-- Every one-round algorithm has monochromatic-edge probability at least `0.23879`. -/ theorem pStar_ge_23879 : ENNReal.ofReal (23879 / 100000 : ℝ) ≤ ClassicalAlgorithm.pStar := by exact ClassicalAlgorithm.le_pStar_of_forall_le (fun alg => Distributed2Coloring.p_ge_23879 alg) /-- An explicit one-round algorithm achieves monochromatic-edge probability at most `0.24118`. -/ theorem pStar_le_24118 : ClassicalAlgorithm.pStar ≤ ENNReal.ofReal (24118 / 100000 : ℝ) := by exact ClassicalAlgorithm.pStar_le_of_exists_le Distributed2Coloring.exists_algorithm_p_le_24118 theorem pStar_lt_24118 : ClassicalAlgorithm.pStar < ENNReal.ofReal (24118 / 100000 : ℝ) := by rcases UpperBound.Recursive3Param.exists_algorithm_p_lt with ⟨alg, hlt⟩ have hpStar_le : ClassicalAlgorithm.pStar ≤ ClassicalAlgorithm.p alg := by exact sInf_le ⟨alg, rfl⟩ exact lt_of_le_of_lt hpStar_le hlt end Distributed2Coloring