Unnumbered Eqs.(29)–(32),(145): Steiner-cut LP identity
OpenQAlgorithms.Papers.Independent.Intakethird100.P35.connectivity_lp_identityLet be a finite undirected graph with finite nonnegative capacities and a set of at least two terminals. An -cut separates at least one pair of terminals, and its capacity is the sum of capacities of its crossing edges, each counted once. Let denote the usual nonnegative oriented-flow LP with net source objective. In the connectivity LP, each unordered pair of terminals has its own full copy of every edge budget and conservation constraints; a common slack is bounded by every pair's net source flow. Then
The cut minimum, pair minimum, and connectivity LP maximum are attained. This expresses Steiner connectivity as a linear program of independent pair tests, including disconnected graphs and zero capacities.
import Definitions.Def_QAlgorithms_Papers_Independent_Intakethird100_P35
noncomputable section
namespace QAlgorithms.Papers.Independent.Intakethird100.P35
open scoped BigOperators
attribute [local instance] Classical.propDecidable Classical.decEq
/-- PDF pp.15,30, (29)--(32),(145): Steiner cut, smallest pair max-flow,
and the independently budgeted connectivity LP have equal values, with both
extrema attained. S is a set of at least two distinct vertices. -/
theorem connectivity_lp_identity {n e : ℕ} (G : EdgeGraph n (Fin e))
(hSimple : IsSimple G) (c : Fin e → ℝ) (hc : ∀ a, 0 ≤ c a)
(S : Finset (Fin n)) (hS : 2 ≤ S.card) :
steinerValue G c S = connectivityValue G c S ∧
(∃ p : TerminalPair S,
singlePairFlowValue G c p.val.1 p.val.2 = connectivityValue G c S ∧
∀ q : TerminalPair S,
singlePairFlowValue G c p.val.1 p.val.2 ≤ singlePairFlowValue G c q.val.1 q.val.2) ∧
(∃ A, SteinerCut S A ∧ cutCapacity G c A = steinerValue G c S ∧
∀ B, SteinerCut S B → cutCapacity G c A ≤ cutCapacity G c B) ∧
(∃ f, ConnectivityFeasible G c S (connectivityValue G c S) f) := by sorry
end QAlgorithms.Papers.Independent.Intakethird100.P35
end
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
For all natural numbers , consider a graph on vertices with edges , distinct designated endpoints for each edge, and the property that edges with identical endpoints in either order are equal; let be real capacities and let be a vertex set with . For a real assignment , with , define
For vertices , call feasible for when all its entries are nonnegative, for each edge, and at every ; let be the supremum of the source outflows of these feasible assignments. Let , and let be the supremum of all real for which there is a family such that each is feasible for and ; these capacity bounds hold separately for each pair, with no capacity sum across pairs. For any vertex set , let be the sum of over edges with exactly one endpoint in , call a terminal-separating cut when there exist with and , and let be the infimum of the capacities of all such cuts. Then all four assertions hold: ; there exists with and for every ; there exists a terminal-separating cut with and for every terminal-separating cut ; and there exists a family of the stated separately feasible assignments with for every . The cardinality hypothesis excludes , but empty edge sets, zero capacities, and disconnected terminal pairs are permitted.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.