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.

50 Chapter 3. <strong>Edge</strong>-<strong>connectivity</strong> augmentation <strong>of</strong> <strong>hypergraphs</strong><br />

3.1.3 Hypergraphs: various objectives<br />

<strong>Edge</strong>-<strong>connectivity</strong> augmentation problems for <strong>hypergraphs</strong> have been less intensively stud-<br />

ied than their graph counterparts. However, some recent results have shown that they<br />

are also worthy <strong>of</strong> interest. In [11], Cheng gave a formula on the minimum number <strong>of</strong><br />

graph edges that can be added to an initial (k − 1)-edge-connected hypergraph so that<br />

the resulting hypergraph is k-edge-connected; Bang-Jensen <strong>and</strong> Jackson [2] extended this<br />

result to the case when the initial hypergraph can be arbitrary. More general frameworks,<br />

involving the covering <strong>of</strong> set functions or <strong>of</strong> families <strong>of</strong> sets, were proposed by Benczúr<br />

<strong>and</strong> Frank [9] <strong>and</strong> Fleiner <strong>and</strong> Jordán [20]. Szigeti [69] solved the local edge-<strong>connectivity</strong><br />

augmentation problem for <strong>hypergraphs</strong> when the aim is to minimize the total size <strong>of</strong> the<br />

added hyperedges.<br />

These results show that the objective <strong>of</strong> minimizing the number <strong>of</strong> new edges, which we<br />

used in the graph case, can be generalized in various directions. In the rest <strong>of</strong> the chapter<br />

we address these possibilities. Of course minimizing the number <strong>of</strong> new hyperedges is in<br />

itself an uninteresting question, since it is always best to use hyperedges containing the<br />

whole ground set. One alternative is to consider instead the total size (the sum <strong>of</strong> the<br />

sizes) <strong>of</strong> the hyperedges; another is to constrain the size <strong>of</strong> the added hyperedges.<br />

In the following sections we cite the known results about these objectives. We also<br />

describe the generalization studied in [9] <strong>and</strong> [69], that involves covering certain types<br />

<strong>of</strong> set functions by a hypergraph. The chapter is concluded by a new result on covering<br />

symmetric crossing supermodular functions by uniform <strong>hypergraphs</strong>, which extends results<br />

in [20] <strong>and</strong> [9].<br />

3.2 Minimum size augmentation<br />

3.2.1 Local edge-<strong>connectivity</strong> requirements<br />

Let us first consider the objective <strong>of</strong> minimizing the sum <strong>of</strong> the sizes <strong>of</strong> the added hyper-<br />

edges, called the total size. It turns out that in some sense this problem is actually easier<br />

then graph edge-<strong>connectivity</strong> augmentation. Local edge-<strong>connectivity</strong> augmentation can be<br />

solved, <strong>and</strong> it does not require tools as sophisticated as Theorem 3.3 <strong>of</strong> Mader.<br />

Let r : V 2 → Z+ be a local edge-<strong>connectivity</strong> requirement function for which r(u, v) =<br />

r(v, u) <strong>and</strong> r(v, v) = 0 for every u, v ∈ V . For a set X ⊆ V , let R(X) := max{r(u, v) :<br />

u ∈ X, v ∈ V − X}. Szigeti [69] proved the following:<br />

Theorem 3.8 (Szigeti [69]). Let H0 = (V, E0) be a hypergraph, <strong>and</strong> r a local edge-<br />

<strong>connectivity</strong> requirement function like above. There is a hypergraph H with hyperedges <strong>of</strong>

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

Saved successfully!

Ooh no, something went wrong!