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.
44 Chapter 2. Submodular functions<br />
Proposition 2.25. Let N = N(W, A1, A2) be a network matrix. Let l : A1 → Z ∪ {−∞},<br />
u : A1 → Z ∪ {∞} be lower <strong>and</strong> upper capacities on A1 such that l ≤ u, <strong>and</strong> let f : A2 →<br />
Z ∪ {−∞}, g : A2 → Z ∪ {∞} be lower <strong>and</strong> upper capacities on A2 such that f ≤ g. Then<br />
the system<br />
is TDI.<br />
{x : A2 → Q : l ≤ Nx ≤ u, f ≤ x ≤ g} (2.25)<br />
To see an example, consider a digraph D = (V, A), <strong>and</strong> a cross-free family F on the<br />
ground set V . Let (T = (W, A1), ϕ) be the tree-representation <strong>of</strong> F. We can associate<br />
a digraph D ′ = (W, A2) to D by taking the edge ϕ(u)ϕ(v) for every edge uv ∈ A. We<br />
get a network matrix N = N(W, A1, A2). The system (2.25) described in Proposition 2.25<br />
corresponds to the following system, if we consider l, u to be given on the sets <strong>of</strong> F, <strong>and</strong><br />
f, g to be given on the edges <strong>of</strong> D:<br />
{x : A → Q : l(Z) ≤ δx(Z) − ϱx(Z) ≤ u(Z) ∀Z ∈ F, f ≤ x ≤ g}, (2.26)<br />
where δx(Z) = <br />
a∈∆ +<br />
D (Z) x(a) <strong>and</strong> ϱx(Z) = <br />
system is TDI.<br />
2.4.2 Submodular flows<br />
a∈∆ −<br />
D<br />
(Z) x(a). So by proposition 2.25, this<br />
To simplify the notations later in the thesis, we describe submodular flows using super-<br />
modular functions, but this is <strong>of</strong> course equivalent to the submodular formulation.<br />
Let D = (V, A) be a digraph, <strong>and</strong> p : 2 V → Z∪{−∞} a crossing supermodular function.<br />
Let furthermore f : A → Z ∪ {−∞}, g : A → Z ∪ {∞} be upper <strong>and</strong> lower capacities on A<br />
such that f ≤ g. For x : A → Q, let φx(Z) := δx(Z) − ϱx(Z), which is a modular function.<br />
The system<br />
{x : A → Q : φx(Z) ≥ p(Z) ∀Z ⊆ V, f ≤ x ≤ g} (2.27)<br />
is called a submodular flow system. A simple example is system (2.26), where p(X) =<br />
max{l(X), −u(V − X)} if X ∈ F + co(F) ,<strong>and</strong> p(X) = −∞ otherwise.<br />
A submodular flow system is one-way if p(X) > −∞ implies that either ϱD(X) = 0<br />
or δD(X) = 0. It is strongly one-way if either ϱD(X) = 0 whenever p(X) > −∞, or<br />
δD(X) = 0 whenever p(X) > −∞.<br />
The fundamental result on submodular flows is the following:<br />
Theorem 2.26 (Edmonds <strong>and</strong> Giles[17]). The system (2.27) is TDI.