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 3.4. Covering by uniform <strong>hypergraphs</strong> 57<br />

every v ∈ V . For such a hyperedge let<br />

m e (v) := m(v) − |e ∩ v|, (3.12)<br />

p e <br />

(p(X) − 1)<br />

(X) :=<br />

+ if e enters X,<br />

(3.13)<br />

p(X) otherwise.<br />

We say that (m e , p e ) is obtained from (m, p) by splitting <strong>of</strong>f the hyperedge e. A splitting-<strong>of</strong>f<br />

operation is feasible if (3.9), (3.10), <strong>and</strong> (3.11) are true for m e <strong>and</strong> p e . It is easy to see that<br />

p e is symmetric <strong>and</strong> positively crossing supermodular; so after the execution <strong>of</strong> a feasible<br />

splitting-<strong>of</strong>f, by induction there exists a ν-uniform hypergraph H ′ with degree vector m e<br />

that covers p e . By adding the hyperedge e to H ′ we obtain a feasible hypergraph H.<br />

The rest <strong>of</strong> the pro<strong>of</strong> consists <strong>of</strong> showing that a feasible splitting-<strong>of</strong>f always exists. We<br />

define the following families <strong>of</strong> sets:<br />

B1 := {∅ = X ⊆ V : m(X) − p(X) ≤ ν − 2, p(Y ) < p(X) ∀Y ⊂ X},<br />

B2 := {X ⊆ V : m(X) − p(X) = ν − 1, p(X) > 0, p(Y ) ≤ p(X) ∀Y ⊂ X},<br />

m(V ) m(V )<br />

B3 := {X ⊆ V : p(X) = , p(Y ) < ∀Y ⊂ X}.<br />

ν<br />

ν<br />

Let e be a ν-hyperedge that can be split <strong>of</strong>f, <strong>and</strong> suppose that m e <strong>and</strong> p e violates (3.9)<br />

because there is a set X entered by e such that p(X) − 1 > m(X) − |e ∩ X|. Notice that<br />

m(X ′ ) − |e ∩ X ′ | ≤ m(X) − |e ∩ X| for every X ′ ⊆ X. Thus there must be a set X ′ ∈ B1<br />

such that X ′ ⊆ X <strong>and</strong> p(X ′ ) − 1 > m(X ′ ) − |e ∩ X ′ |. Similarly, if m e <strong>and</strong> p e violates (3.9)<br />

because there is a set X such that e ⊆ X <strong>and</strong> p(X) = m(X) − (ν − 1), then either there is<br />

a set X ′ ∈ B2 such that X ′ ⊆ X <strong>and</strong> e ⊆ X ′ , or there is a set X ′ ∈ B1 such that X ′ ⊆ X<br />

<strong>and</strong> p(X ′ ) − 1 > m(X ′ ) − |e ∩ X ′ |.<br />

A similar argument shows that if m e <strong>and</strong> p e violate (3.9) then there is a set X ∈ B3 such<br />

that e ⊆ V − X. We obtained the following:<br />

Claim 3.15. The inequalities (3.9) <strong>and</strong> (3.10) hold for m e <strong>and</strong> p e if <strong>and</strong> only if<br />

|e ∩ X| ≤ m(X) − p(X) + 1 for every X ∈ B1, (3.14)<br />

|e ∩ X| ≤ ν − 1 for every X ∈ B2, (3.15)<br />

|e ∩ X| ≥ 1 for every X ∈ B3. (3.16)<br />

Furthermore, if m e (X) < p e (X) for some X, then there is a subset X ′ <strong>of</strong> X such that<br />

either X ′ ∈ B1 <strong>and</strong> it violates (3.14), or X ′ ∈ B2 <strong>and</strong> it violates (3.15).<br />

In order to formulate necessary <strong>and</strong> sufficient conditions for m e <strong>and</strong> p e to satisfy (3.11),<br />

we call a p-full partition {V1, . . . , Vl} critical if<br />

l − 1<br />

ν − 1<br />

> m(V )<br />

ν<br />

− 1.

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

Saved successfully!

Ooh no, something went wrong!