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 6.4. Directed network design problems with orientation constraints 105<br />

v1 a1<br />

v2 v3<br />

v1 v2 v3 v4 v5 v6<br />

v4<br />

e1<br />

a2<br />

v5<br />

e2<br />

v6<br />

Figure 6.2: Transformation into a submodular flow problem in Theorem 6.13<br />

correspond to the hyperedges <strong>and</strong> hyperarcs in E ∪ A; we denote a node corresponding to<br />

a hyperedge or hyperarc e by ue. The edge set B contains an edge from the head <strong>of</strong> e to ue<br />

(with edge cost equal to the cost <strong>of</strong> e) if e is a hyperarc; if e is a hyperedge, then B contains<br />

edges from every node <strong>of</strong> e to ue (each with cost equal to the cost <strong>of</strong> the corresponding<br />

orientation <strong>of</strong> e). A set function q is defined on the ground set U ∪ V as follows:<br />

⎧<br />

⎪⎨ p(X ∩ V ) if ue ∈ X implies that the nodes <strong>of</strong> e are in X,<br />

q(X) := −1<br />

⎪⎩<br />

−∞<br />

if X = {ue} for some e ∈ E,<br />

otherwise.<br />

Figure 6.2 shows this construction. Let f ≡ 0 be the lower capacity <strong>of</strong> the edges <strong>and</strong><br />

g ≡ 1 the upper capacity. It is easy to see that if p is intersecting supermodular, then q is<br />

crossing supermodular. So Theorem 2.28 implies that the problem <strong>of</strong> finding a minimum<br />

ua1<br />

cost <strong>directed</strong> subgraph D ′ = (U, V ; B ′ ) <strong>of</strong> D that satisfies<br />

δD ′(X) − ϱD ′(X) ≥ q(X) for every X ⊆ U ∪ V (6.6)<br />

can be solved in polynomial time (note that this is a one-way submodular flow system).<br />

Since q({ue}) = −1 if e ∈ E, ue is the head <strong>of</strong> at most one edge <strong>of</strong> D ′ . Thus the subgraph<br />

D ′ corresponds to an oriented sub-hypergraph M ′ <strong>of</strong> M. It is easy to check that M ′ covers<br />

the requirement function p if <strong>and</strong> only if D ′ satisfies (6.6).<br />

6.4.2 TDI property<br />

The above construction shows that when the requirement function is intersecting supermod-<br />

ular, the <strong>directed</strong> network design problem with orientation constraints can be transformed<br />

into a submodular flow problem which is TDI. We now prove that the linear system that<br />

we naturally associate to the original problem is also TDI, even for positively intersecting<br />

supermodular requirement functions. To formulate the appropriate linear program, the<br />

hypergraph analogue <strong>of</strong> orientation constraints must be defined. The elements <strong>of</strong> a set A ′<br />

ua2<br />

ue1<br />

ue2<br />

V<br />

U

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

Saved successfully!

Ooh no, something went wrong!