14.02.2013 Views

Gruber P. Convex and Discrete Geometry

Gruber P. Convex and Discrete Geometry

Gruber P. Convex and Discrete Geometry

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

58 <strong>Convex</strong> Bodies<br />

v<br />

HC (u, h(u))<br />

(v, h(v))<br />

o<br />

−1<br />

R<br />

x<br />

(C, −1)<br />

(x, −1)<br />

(u, h(u))<br />

Fig. 4.2. A homogeneous convex function is a support function<br />

4.2 Separation <strong>and</strong> Oracles<br />

Separation of convex sets, in particular of convex polytopes <strong>and</strong> convex polyhedra,<br />

plays an important role in optimization <strong>and</strong> convex analysis.<br />

In this section, we show a st<strong>and</strong>ard separation theorem which for most applications<br />

is sufficient, <strong>and</strong> describe oracles to specify convex bodies.<br />

For more information on separation <strong>and</strong> oracles, see Stoer <strong>and</strong> Witzgall [970],<br />

Hiriart-Urruty <strong>and</strong> Lemaréchal [505], <strong>and</strong> Grötschel, Lovász <strong>and</strong> Schrijver [409].<br />

Separation <strong>and</strong> Strong Separation<br />

<strong>Convex</strong> sets C <strong>and</strong> D in E d are separated if there is a hyperplane H such that<br />

C ⊆ H − <strong>and</strong> D ⊆ H + or vice versa, where H + <strong>and</strong> H − are the closed halfspaces<br />

determined by H. Then H is called a separating hyperplane of C <strong>and</strong> D. The<br />

sets C <strong>and</strong> D are strongly separated if there is a closed slab S with int S �= ∅such<br />

that C ⊆ S − <strong>and</strong> D ⊆ S + or vice versa. Here S − <strong>and</strong> S + are the closed halfspaces,<br />

determined by the two boundary hyperplanes of S, not containing S. We call S a<br />

separating slab of C <strong>and</strong> D.<br />

In the following, two separation results will be presented. The first result is a<br />

simple yet useful tool.<br />

Proposition 4.1. Let C, D be convex sets in E d . Then the following statements are<br />

equivalent:<br />

(i) C <strong>and</strong> D are separated, respectively, strongly separated.<br />

(ii) The convex set C − D = {x − y : x ∈ C, y ∈ D} <strong>and</strong> {o} are separated,<br />

respectively, strongly separated.<br />

Proof. We consider only the case of strong separation, the case of separation is<br />

similar, but simpler.<br />

u<br />

E d

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

Saved successfully!

Ooh no, something went wrong!