Global shortest-path greedy is within a factor of four
OpenQAlgorithms.Papers.Independent.Intakethird100.P45.greedyWithinFourConsider a finite square-grid network with distinct terminals in the paper's one-slot heralded-loss model. Each physical edge independently supplies at most one perfect EPR pair with probability . After seeing all external-success flags, a schedule selects disjoint memory pairs for simultaneous BSMs, whose independent success probability is . The terminals retain their delivered halves. Let be the expected delivered ebits per slot under repeated shortest-path selection and edge deletion, using a fixed shortest-path tie choice. Let be the optimum expected rate over all such one-slot pairwise-BSM schedules that observe external flags but cannot see future BSM outcomes. Then
The rates are constructed by averaging the conditional path-sum yields over independent external-success subgraphs; disconnected slots contribute zero. This is a restricted-protocol guarantee, rather than a statement about unrestricted quantum-network capacity.
import Definitions.Def_QAlgorithms_Papers_Independent_Intakethird100_P45
namespace QAlgorithms.Papers.Independent.Intakethird100.P45
open scoped BigOperators
noncomputable section
/-- PDF pp.5–7: actual shortest-path greedy achieves at least one quarter
of the actual global optimum, and no more than the optimum. Both rates and
all pairings are constructed above. Disconnected slots contribute zero.
Endpoints `p,q=0,1` are allowed; a direct edge has zero BSMs, even at `q=0`. -/
theorem greedyWithinFour (N : ℕ) (G : GridNetwork N) (a b : Fin N)
(hab : a ≠ b) (p q : ℝ) (hp0 : 0 ≤ p) (hp1 : p ≤ 1)
(hq0 : 0 ≤ q) (hq1 : q ≤ 1) :
optimalGlobalRate G a b p q / 4 ≤ greedyRate G a b p q ∧
greedyRate G a b p q ≤ optimalGlobalRate G a b p q := by sorry
end
end QAlgorithms.Papers.Independent.Intakethird100.P45
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
For every natural number , every grid network given by an injective placement of , every pair of distinct vertices , and every pair of real numbers satisfying and , the following two inequalities hold:
Here consists of pairs with whose positions have Manhattan distance . An encoded route consists of a length and a vertex sequence ; it is admissible in an edge set from to when its endpoints are , its canonical pairs are distinct members of , and no intermediate vertex equals either endpoint. Other vertex repetitions are permitted. A feasible family is a finite set of such routes whose edge sets are disjoint for every two distinct members, , and is the maximum of and all values for feasible families in . The family is obtained by starting with and available steps: if there are no steps or no admissible routes, return the empty set; otherwise select a minimum-length admissible route by a fixed classical choice, delete its edge set, recursively continue with one fewer step, and insert that route into the resulting finite set. The outer sums include all subsets of , including the empty subset; empty family sums are , and every zeroth power, including , equals . Thus the stated parameter ranges include , , , and . For or , there are no distinct endpoint choices satisfying the hypotheses.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.