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

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

34 Chapter 2. Submodular functions<br />

2.2.4 Intersecting <strong>and</strong> crossing supermodularity<br />

For set functions that appear in <strong>connectivity</strong> problems, the supermodular inequality (2.2)<br />

<strong>of</strong>ten does not hold for every pair <strong>of</strong> sets. For example, given a hypergraph H0 <strong>and</strong> a<br />

positive integer k, let p(X) := k − dH0(X) if ∅ = X ⊂ V , <strong>and</strong> p(∅) := p(V ) := 0. For a<br />

hypergraph H, H0 + H is k-edge-connected if <strong>and</strong> only if H covers p. The set function p<br />

is not necessarily supermodular on non-crossing sets.<br />

Another example is for a given <strong>directed</strong> hypergraph D0 with a fixed root node s: let<br />

p(X) := k −ϱD0(X) is ∅ = X ⊆ V −s, <strong>and</strong> p(X) := 0 otherwise. For a <strong>directed</strong> hypergraph<br />

D, D0 + D is k-rooted-connected from s if <strong>and</strong> only if D covers p. Here p is supermodular<br />

on intersecting sets.<br />

This shows that in many cases the requirement <strong>of</strong> full supermodularity must be relaxed.<br />

A set function p : 2 V → Z ∪ {−∞} is crossing supermodular if (2.2) holds whenever X <strong>and</strong><br />

Y are crossing. The set function p is intersecting supermodular if (2.2) holds whenever X<br />

<strong>and</strong> Y are intersecting.<br />

For crossing supermodular functions, the following theorem <strong>of</strong> Fujishige characterizes<br />

the non emptiness <strong>of</strong> B(p).<br />

Theorem 2.11 (Fujishige [38]). Let p : 2 V → Z ∪ {−∞} be a crossing supermodular<br />

function. Then B(p) is nonempty if <strong>and</strong> only if<br />

t<br />

p(Xi) ≤ p(V ),<br />

i=1<br />

t<br />

p(V − Xi) ≤ (t − 1)p(V )<br />

i=1<br />

both hold for every partition {X1, X2, . . . , Xt} <strong>of</strong> V . Furthermore, if B(p) is non-empty,<br />

then it is a base polyhedron.<br />

If p is intersecting supermodular, then the non-emptiness condition reduces to the fol-<br />

lowing:<br />

Proposition 2.12. Let p : 2 V → Z ∪ {−∞} be an intersecting supermodular function.<br />

Then B(p) is nonempty if <strong>and</strong> only if<br />

t<br />

p(Xi) ≤ p(V )<br />

i=1<br />

holds for every partition {X1, X2, . . . , Xt} <strong>of</strong> V . Furthermore, if B(p) is non-empty, then<br />

it is a base polyhedron.

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

Saved successfully!

Ooh no, something went wrong!