Lemma 4.18, reduced case (Fig. 2, Cases 1–3) — expected make-span at most 7/4 · t_opt
ProvedAlgMechDesign.Randomized.reduced_case_boundinequalityp2o-batch-p200bp2o-gran-per-chapterp2o-plan-paperp2o-v1randomized-mechanismscheduling
Let and . Consider the four-task instance of Fig. 2 for two agents: task with times , with , with and with for agents . The mechanism gives to agent 1 and gives each of to either agent with probability ; the optimal allocation gives to agent 1 and to agent 2. If both agents finish at the same time under the optimal allocation, , then the expected make-span over the four equally likely allocations satisfies
This is the case to which the paper reduces Lemma 4.18; zero times represent instances with fewer than four tasks.
Formalization Note The three cases of the paper are combined into one inequality with explicit maxima, with substituted.
Preamble
import Mathlib
Formal statement
namespace AlgMechDesign.Randomized
/-- The reduced case of Lemma 4.18 (Fig. 2, Cases 1–3, pp. 184–185), with `β = 4/3`: tasks
`k₁, k₂` (times `(a, βa)` and `(b, βb)`) go to agent 1, while `l₁` (times `(c, βc)`) and `l₂`
(times `(βd, d)`) are each allocated to either agent with probability `1/2`. If opt's two
finishing times agree, `a + c = βb + d`, the expected make-span is at most `7/4 · (a + c)`. -/
theorem reduced_case_bound (a b c d : ℝ) (ha : 0 ≤ a) (hb : 0 ≤ b) (hc : 0 ≤ c) (hd : 0 ≤ d)
(hopt : a + c = 4 / 3 * b + d) :
(1 / 4 : ℝ) * (max (a + b + c + 4 / 3 * d) 0 + max (a + b + c) d +
max (a + b + 4 / 3 * d) (4 / 3 * c) + max (a + b) (4 / 3 * c + d)) ≤
7 / 4 * (a + c) := by sorry
end AlgMechDesign.Randomized
Source
Nisan, Ronen, Algorithmic Mechanism Design, Games Econ. Behav. 35, 2001, pp. 184–185, proof of Lemma 4.18, Fig. 2 and Cases 1–3
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.