Theorem I.2 — randomized double greedy achieves half the optimum
ProvedDoubleGreedyUSM.Randomized.randomized_usm_halfapproximation-algorithmsp2o-batch-pfp1ap2o-gran-per-chapterp2o-plan-paperp2o-v1randomized-algorithmssubmodular-functions
Let be a submodular set function on a finite ground set, and let be any order of its elements. Run the randomized double-greedy Algorithm 2 and let be its output. If , then
Thus the algorithm attains the paper's one-half approximation ratio in expectation, including when the optimum is zero.
Formalization Note The theorem concerns the specific Algorithm 2 law and every enumeration of the ground set. It encodes the approximation inequality, while the paper's linear-time and value-oracle complexity claims are outside the Lean statement. The output identity is stated separately in the endpoint milestone.
Preamble
import Mathlib import Definitions.Def_NonmonotoneSubmod_Shared_Submodular import Definitions.Def_NonmonotoneSubmod_Shared_OPT import Definitions.Def_DoubleGreedyUSM_Randomized_Algorithm2
Formal statement
namespace DoubleGreedyUSM.Randomized
/-- Theorem I.2 (PDF p. 2): Algorithm 2 has expected output at least half the optimum. -/
theorem randomized_usm_half {X : Type} [Fintype X] [DecidableEq X]
(f : Finset X → ℝ) (hf : NonmonotoneSubmod.Shared.Submodular f)
(hf0 : ∀ S : Finset X, 0 ≤ f S)
(l : List X) (hl : l.Nodup) (hcov : ∀ x : X, x ∈ l) :
NonmonotoneSubmod.Shared.OPT f ≤
2 * expect (state f l l.length) (fun s => f s.1) := by sorry
end DoubleGreedyUSM.Randomized
Source
Buchbinder, Feldman, Naor, Schwartz, A Tight Linear Time (1/2)-Approximation for Unconstrained Submodular Maximization, FOCS 2012 version, Theorem I.2 (PDF p. 2), proof (PDF p. 5)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.