Edge-connectivity of undirected and directed hypergraphs
Edge-connectivity of undirected and directed hypergraphs
Edge-connectivity of undirected and directed hypergraphs
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Section 3.4. Covering by uniform <strong>hypergraphs</strong> 67<br />
We choose a hyperedge e that can be feasibly split <strong>of</strong>f by the method described in the<br />
pro<strong>of</strong> <strong>of</strong> Theorem 3.14. If m(v) ≥ ν for every v ∈ e then the following extended splitting-<strong>of</strong>f<br />
operation may be used:<br />
(⋆) We determine the maximal µ for which e can be feasibly split <strong>of</strong>f µ times <strong>and</strong> µχe(v) ≤<br />
m(v) − ν + 1 for every v ∈ e.<br />
Claim 3.31. An extended splitting-<strong>of</strong>f can be executed in polynomial time.<br />
Pro<strong>of</strong>. The degree restriction ensures that condition (3.17) is indifferent when calculating<br />
the number <strong>of</strong> feasible splitting-<strong>of</strong>f operations, since the auxiliary graph is the same before<br />
each splitting-<strong>of</strong>f. So it suffices to determine the maximal value µ for which the following<br />
hold:<br />
µχe(v) ≤ m(v) − (ν − 1) for every v ∈ e, (3.27)<br />
µ(|e ∩ X| − 1) ≤ m(X) − p(X) for every X entered by e, (3.28)<br />
µν ≤ m(X) − p(X) if e ⊆ X, (3.29)<br />
µ ≤<br />
m(V )<br />
ν<br />
− p(X) if e ⊆ X. (3.30)<br />
The maximal µ for which (3.29) <strong>and</strong> (3.30) hold can be determined by the maximizing<br />
oracle. As for (3.28), it suffices to check its validity on the family<br />
{X ⊆ V : p(Y ) < p(X) ∀Y ⊂ X, e enters X}.<br />
This family can be determined in polynomial time since it is laminar.<br />
The extended splitting-<strong>of</strong>f operation is not used if m(v) ≤ ν − 1 for some v ∈ e; a single<br />
splitting-<strong>of</strong>f is executed instead. But the number such splitting-<strong>of</strong>f operations is polynomial<br />
(we can assume that ν ≤ |V |).<br />
We have to prove that the number <strong>of</strong> extended splitting-<strong>of</strong>f operations is also polynomial.<br />
First, observe that no set is deleted from B1 during an extended splitting-<strong>of</strong>f. Indeed,<br />
X ∈ B1 is deleted during a single splitting-<strong>of</strong>f <strong>of</strong> e if there is Y ⊂ X for which p(Y ) < p(X)<br />
<strong>and</strong> p e (Y ) = p e (X); but then |e ∩ (X − Y )| > 0, so m e (X − Y ) ≥ ν − 1, implying that<br />
m e (Y ) − p e (Y ) ≤ m e (X) − (ν − 1) − p e (X) < 0, which is impossible. Since B1 is always<br />
laminar, this implies that in a sequence <strong>of</strong> consecutive extended splitting-<strong>of</strong>f operations,<br />
B1 can change only polynomially many times. It is easy to see that B3 also can change<br />
only polynomially many times. So it is enough to prove that if B1 <strong>and</strong> B3 do not change,<br />
then only polynomially many consecutive extended splittings are possible. But this follows<br />
from the fact that ν − 1 nodes are the same in the hyperedges that are split <strong>of</strong>f.