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.

Section 5.2. Tree-reducible <strong>hypergraphs</strong> 91<br />

There is a difference in the structural properties <strong>of</strong> k-partition-connected <strong>hypergraphs</strong>,<br />

<strong>and</strong> (k, l)-partition-connected <strong>hypergraphs</strong> in general. Namely, Theorem 5.7 implies that a<br />

k-partition-connected hypergraph is always reducible to a k-partition-connected graph, in<br />

the sense used in the definition <strong>of</strong> forest-reducibility. However, a (k, l)-partition-connected<br />

hypergraph is not necessarily reducible to a (k, l)-partition-connected graph. For exam-<br />

ple, the hypercircuits on Figure 5.2 are not reducible (<strong>of</strong> course there are also reducible<br />

hypercircuits).<br />

If we want to prove someone that a given hypergraph is k-partition-connected, we can do<br />

so by showing its reduction to a k-partition-connected graph, which contains k edge-disjoint<br />

spanning trees. But how can we prove that a given hypergraph is (k, l)-partition-connected?<br />

An answer to this question will be given in Chapter 6.<br />

5.2.5 Covering by hyperforests<br />

The problems discussed so far in this chapter were generalizations <strong>of</strong> problems that involve<br />

the packing <strong>of</strong> trees. It may be asked how problems related to covering with forests<br />

generalize to <strong>hypergraphs</strong>. An extension <strong>of</strong> Theorem 1.6 <strong>of</strong> Nash-Williams on partitioning<br />

a graph into k forests easily follows from Theorem 5.6.<br />

Theorem 5.9. The edge-set E <strong>of</strong> a hypergraph H = (V, E) can be decomposed into k<br />

hyperforests if <strong>and</strong> only if<br />

holds for every nonempty subset X <strong>of</strong> V .<br />

iH(X) ≤ k(|X| − 1) (5.5)<br />

Pro<strong>of</strong>. Since one hyperforest may contain at most |X| − 1 hyperedges induced by X, (5.5)<br />

is necessary. To see the sufficiency, let F := {V1, . . . , Vt} be a partition <strong>of</strong> V . By (5.5) each<br />

member Vi induces at most k(|Vi|−1) hyperedges, so the number eH(F) <strong>of</strong> cross-hyperedges<br />

is at least |E| − t<br />

j=1 k(|Vj| − 1) = |E| − k(|V | − t). Therefore k(|V | − |F|) + eH(F) ≥ |E|<br />

holds for every partition F <strong>of</strong> V , which means that E is independent in the matroid MkH<br />

by (5.4), <strong>and</strong> H decomposes into k hyperforests.<br />

The problem <strong>of</strong> <strong>connectivity</strong> augmentation with respect to (k, l)-partition-connectivi-<br />

ty will be discussed in Chapter 7. Here we address the much easier analogous problem<br />

concerning covering with hyperforests: add hyperedges to a given hypergraph H0 = (V, E0)<br />

such that the resulting hypergraph can still be decomposed into k hyperforests. It is a<br />

simple observation that even the maximum weight problem is solvable (assuming that<br />

the number <strong>of</strong> hyperedges with positive weight is polynomial), by the matroid techniques

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

Saved successfully!

Ooh no, something went wrong!