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.

Section 4.3. Covering supermodular functions by <strong>directed</strong> <strong>hypergraphs</strong> 75<br />

the resulting (r, 1)-hyperarc. By induction, there is a <strong>directed</strong> (r, 1)-hypergraph D ′ that<br />

satisfies the conditions given by m a i , m a o <strong>and</strong> p a . The <strong>directed</strong> hypergraph obtained by<br />

adding a to D ′ satisfies the conditions <strong>of</strong> Theorem 4.5.<br />

In fact, a bit more general theorem can also be proved by the same method (this was<br />

also remarked by Berg, Jackson, <strong>and</strong> Jordán for the k-edge-<strong>connectivity</strong> case):<br />

Theorem 4.7. Let p : 2 V → Z+ be a positively crossing supermodular set function, mi :<br />

V → Z+ <strong>and</strong> mo : V → Z+ degree specifications such that mi(V ) ≤ mo(V ) <strong>and</strong> (r −<br />

1)mi(V ) < mo(V ) ≤ rmi(V ) for some positive integer r, <strong>and</strong><br />

mi(X) ≥ p(X) for every X ⊆ V , (4.5)<br />

mo(V − X) ≥ p(X) for every X ⊆ V . (4.6)<br />

Then there is a <strong>directed</strong> hypergraph D consisting <strong>of</strong> (r, 1)-hyperedges <strong>and</strong> (r − 1, 1)-hyper-<br />

edges such that δD(v) = mo(v) <strong>and</strong> ϱD(v) = mi(v) for every v ∈ V , <strong>and</strong><br />

ϱD(X) ≥ p(X) for every X ⊆ V .<br />

Pro<strong>of</strong>. By Theorem 4.3 there is a complete splitting-<strong>of</strong>f, so there must be an (r1, 1)-hyperarc<br />

a that can be feasibly split <strong>of</strong>f for some r1 ≥ r, since (r − 1)mi(V ) < mo(V ). By Theorem<br />

4.3 there is also an (r2, 1)-hyperarc for some r2 ≤ r with head h(a) that can be feasibly<br />

split <strong>of</strong>f, since mo(V ) ≤ rmi(V ). So Lemma 4.6 implies that there is an (r, 1)-hyperarc<br />

that can be feasibly split <strong>of</strong>f. We can continue to split <strong>of</strong>f (r, 1)-hyperarcs as long as<br />

(r − 1)m ′ i(V ) < m ′ o(V ) ≤ rm ′ i(V ) holds for the modified degree-specifications. It is easy<br />

to see that the first time this does not hold is when (r − 1)m ′ i(V ) = m ′ o(V ). But then by<br />

Theorem 4.5 we can finish the complete splitting-<strong>of</strong>f by splitting <strong>of</strong>f (r−1, 1)-hyperarcs.<br />

4.3 Covering supermodular functions by <strong>directed</strong> hy-<br />

pergraphs<br />

Let p : 2 V → Z+ be a positively crossing supermodular set function. When covering p with<br />

a <strong>directed</strong> hypergraph, we can have various objectives just as in the case <strong>of</strong> <strong>un<strong>directed</strong></strong><br />

<strong>hypergraphs</strong>: we can minimize the total size <strong>of</strong> the added hyperarcs, or we can cover p<br />

with an (r, 1)-uniform <strong>directed</strong> hypergraph <strong>of</strong> minimum cardinality.

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

Saved successfully!

Ooh no, something went wrong!