22.08.2013 Views

Edge-connectivity of undirected and directed hypergraphs

Edge-connectivity of undirected and directed hypergraphs

Edge-connectivity of undirected and directed hypergraphs

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

108 Chapter 6. Hypergraph orientation<br />

The columns <strong>of</strong> the matrix will represent a set A ′ 2 <strong>of</strong> <strong>directed</strong> edges, with a one-to-one<br />

correspondence between the hyperarcs in A <strong>and</strong> the edges in A ′ 2. To a hyperarc a ∈ Ai,<br />

assign an edge wih(a). To hyperarcs a ∈ A not appearing in any Ai, assign an edge wZh(a),<br />

where Z is the node set <strong>of</strong> a.<br />

Let N denote the network matrix given by the above network (W ′ ; A ′ 1, A ′ 2). Then by<br />

Proposition 2.25 the system<br />

{x : A ′ 2 → Q : l ′ ≤ Nx ≤ u ′ , f ≤ x ≤ g}<br />

is TDI. Moreover, by the one-to-one correspondence between the edges in A ′ 2 <strong>and</strong> the<br />

hyperarcs in A, this system is equivalent to the system (S ′ ). This implies that (S ′ ) has an<br />

integral dual optimal solution, which in turn is an optimal dual solution for (S).<br />

6.4.3 Min-max theorems<br />

Theorem 6.14 implies that the polyhedron described by (S) is integral, <strong>and</strong> for every integer<br />

cost function there exists an integral optimal dual solution where the family <strong>of</strong> the sets<br />

with positive dual variable is laminar. This allows us to formulate fairly friendly new<br />

min-max formulas for some graph problems. For example, what is the maximum number<br />

<strong>of</strong> <strong>un<strong>directed</strong></strong> edges, or the maximum number <strong>of</strong> <strong>directed</strong> edges, that can be removed<br />

from a mixed graph such that the obtained subgraph has an orientation covering a given<br />

intersecting supermodular set function p? The following corollary describes a min-max<br />

formula that involves both <strong>of</strong> these problems.<br />

Corollary 6.15. Let G = (V ; E, A) be a mixed graph (where E is the set <strong>of</strong> <strong>un<strong>directed</strong></strong><br />

edges <strong>and</strong> A is the set <strong>of</strong> <strong>directed</strong> edges). Let c : E ∪ A → {0, 1} be a cost function, <strong>and</strong><br />

p : 2 V → Z+ a positively intersecting supermodular set function. Then the minimum cost<br />

<strong>of</strong> a subgraph that has an orientation covering p equals<br />

<br />

<br />

p(X) − eE(F) − <br />

<br />

ϱA(X) + µ(F) , (6.14)<br />

max<br />

F laminar<br />

X∈F<br />

where µ(F) is the sum <strong>of</strong> the costs <strong>of</strong> the <strong>un<strong>directed</strong></strong> <strong>and</strong> <strong>directed</strong> edges that enter at least<br />

one member <strong>of</strong> F.<br />

Pro<strong>of</strong>. To formulate the problem in the terms <strong>of</strong> Theorem 6.14, let the <strong>directed</strong> hypergraph<br />

D = (V, A) be the digraph obtained from G by replacing the <strong>un<strong>directed</strong></strong> edges <strong>of</strong> G by a<br />

pair <strong>of</strong> oppositely <strong>directed</strong> edges, <strong>and</strong> let us assign an orientation constraint to every such<br />

pair, with bounds li := 0 <strong>and</strong> ui := 1. Let the cost <strong>of</strong> the edges in a pair be the cost <strong>of</strong> the<br />

X∈F

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!