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.
38 Chapter 2. Submodular functions<br />
We call a set X significant if p(X) = p ∗ (X), <strong>and</strong> tight if p ∗ (X) + p ∗ (V − X) = 0. We<br />
also introduce for X, Y ⊆ V the notation<br />
p ∗ <br />
<br />
(X, Y ) := max p(Z) : F is an (X, Y )-composition . (2.18)<br />
Z∈F<br />
This function clearly has the following properties:<br />
p ∗ (X, Y ) ≥ max{p ∗ (X) + p ∗ (Y ), p ∗ (X ∩ Y ) + p ∗ (X ∪ Y )}, (2.19)<br />
p ∗ (Y ) ≥ p ∗ (X, Y ) + p ∗ (V − X). (2.20)<br />
The main result <strong>of</strong> this section is the following theorem:<br />
Theorem 2.17. If p(X) + p(Y ) ≤ p ∗ (X ∩ Y ) + p ∗ (X ∪ Y ) holds whenever X <strong>and</strong> Y are<br />
significant <strong>and</strong> crossing, then p ∗ is fully supermodular.<br />
Observe that if X <strong>and</strong> Y are not crossing, then p ∗ (X) + p ∗ (Y ) ≤ p ∗ (X ∩ Y ) + p ∗ (X ∪ Y )<br />
automatically holds. So the condition <strong>of</strong> the theorem always holds for non-crossing X <strong>and</strong><br />
Y . Theorem 2.17 implies Proposition 2.15 (it is easy to prove using the uncrossing technique<br />
that p ↑ (X) = p ∗ (X) for every X ⊆ V if p is crossing supermodular <strong>and</strong> p ∗ (V ) = 0).<br />
Furthermore, it shows that instead <strong>of</strong> crossing supermodularity, it is sufficient to require<br />
p(X) + p(Y ) ≤ p ↑ (X ∩ Y ) + p ↑ (X ∪ Y ) for every crossing X, Y .<br />
For technical reasons we will prove a theorem that is a bit more general but less elegant<br />
than Theorem 2.17. We say that a family G generates p∗ if<br />
p ∗ <br />
<br />
(X) = max p ∗ <br />
(Z) : F is a composition <strong>of</strong> X, <strong>and</strong> its members are in G<br />
Z∈F<br />
for every X ⊆ V . Obviously the family <strong>of</strong> significant sets generates p ∗ .<br />
Theorem 2.18. Suppose that a family G generates p ∗ , <strong>and</strong> p ∗ (X, Y ) = p ∗ (X ∩Y )+p ∗ (X ∪<br />
Y ) holds for every pair X, Y ∈ G for which p ∗ (X, Y ) = p ∗ (X) + p ∗ (Y ). Then p ∗ is fully<br />
supermodular.<br />
Pro<strong>of</strong>. The main difficulty <strong>of</strong> the pro<strong>of</strong> is that the composition F <strong>of</strong> X that defines p ∗ (X)<br />
in (2.16) can not always be cross-free (an example for this will be presented at the end<br />
<strong>of</strong> the pro<strong>of</strong>). We will show however that the family F can be assumed to be cross-free<br />
restricted to minimal tight sets. First we prove with the help <strong>of</strong> a few preliminary claims<br />
that p ∗ (X, Y ) = p ∗ (X ∩ Y ) + p ∗ (X ∪ Y ) is true for every tight X <strong>and</strong> arbitrary Y .<br />
Claim 2.19. If X is tight, then p ∗ (X, Y ) = p ∗ (X) + p ∗ (Y ) for every Y ⊆ V .