Edge-connectivity of undirected and directed hypergraphs
Edge-connectivity of undirected and directed hypergraphs
Edge-connectivity of undirected and directed hypergraphs
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Section 4.2. Splitting-<strong>of</strong>f in <strong>directed</strong> <strong>hypergraphs</strong> 71<br />
4.2 Splitting-<strong>of</strong>f in <strong>directed</strong> <strong>hypergraphs</strong><br />
In [7], Berg, Jackson <strong>and</strong> Jordán proved an interesting splitting-<strong>of</strong>f theorem for <strong>directed</strong><br />
<strong>hypergraphs</strong>, which led to a solution for the problem <strong>of</strong> <strong>directed</strong> hypergraph k-edge-<br />
<strong>connectivity</strong> augmentation by uniform hyperarcs.<br />
In this section, we show that their edge-splitting result can be formulated in a more<br />
general form (using essentially the same pro<strong>of</strong>). The result gives a method for solving a<br />
broader class <strong>of</strong> <strong>un<strong>directed</strong></strong> <strong>and</strong> <strong>directed</strong> augmentation problems. The notion <strong>of</strong> splitting-<strong>of</strong>f<br />
is used here in an abstract sense, similarly to Theorem 4.5 or Theorem 3.14.<br />
4.2.1 Splitting <strong>of</strong>f a single hyperarc<br />
Let p : 2 V → Z+ be a positively crossing supermodular set function. Let furthermore<br />
mi : V → Z+ be an indegree-specification <strong>and</strong> mo : V → Z+ an outdegree-specification for<br />
which mi(V ) ≤ mo(V ). Suppose that mi(X) ≥ p(X) <strong>and</strong> mo(V − X) ≥ p(X) for every<br />
X ⊆ V . We define the splitting-<strong>of</strong>f operation analogously to the <strong>un<strong>directed</strong></strong> definition in<br />
the pro<strong>of</strong> <strong>of</strong> Theorem 3.14.<br />
A hyperarc a can be split <strong>of</strong>f from (p, mi, mo) if mi(h(a)) > 0 <strong>and</strong> χt(a)(v) ≤ mo(v) for<br />
every v ∈ V . For such a hyperarc let<br />
m a ⎧<br />
⎨mi(v)<br />
− 1 if v = h(a),<br />
i (v) :=<br />
⎩mi(v)<br />
otherwise,<br />
m a o(v) :=mo(v) − χt(a)(v),<br />
p a ⎧<br />
⎨(p(X)<br />
− 1)<br />
(X) :=<br />
⎩<br />
+ if a enters X,<br />
p(X) otherwise.<br />
The splitting-<strong>of</strong>f operation is feasible if m a i (X) ≥ p a (X) <strong>and</strong> m a o(V − X) ≥ p a (X) for<br />
every X ⊆ V . The operation is called a feasible (r, 1)-splitting if |a| = r + 1. It is easy to<br />
check using (1.4) that p a is positively crossing supermodular.<br />
To motivate the name “splitting-<strong>of</strong>f”, we may observe that if we add a new node z to<br />
V , <strong>and</strong> draw mi(v) digraph edges from z to each node v ∈ V , <strong>and</strong> mo(v) edges from each<br />
node v ∈ V to z, then the above defined splitting-<strong>of</strong>f operation corresponds to splitting-<br />
<strong>of</strong>f in this digraph D. If in addition p is defined as p(X) = (k − ϱD0(X)) + for some<br />
digraph D0, a feasible splitting-<strong>of</strong>f corresponds to a splitting-<strong>of</strong>f in D0 + D that preserves<br />
k-edge-<strong>connectivity</strong> in V (see Figure 4.1).<br />
The following theorem describes conditions when a feasible splitting-<strong>of</strong>f is available.