10.03.2015 Views

v2009.01.01 - Convex Optimization

v2009.01.01 - Convex Optimization

v2009.01.01 - Convex Optimization

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.

4.2. FRAMEWORK 261<br />

When equality is attained in (610)<br />

rankX ⋆ + rankS ⋆ = n (612)<br />

there are no coinciding main diagonal zeros in Λ X ⋆Λ S ⋆ , and so we have what<br />

is called strict complementarity. 4.11 Logically it follows that a necessary and<br />

sufficient condition for strict complementarity of an optimal primal and dual<br />

solution is<br />

X ⋆ + S ⋆ ≻ 0 (613)<br />

4.2.3.1 solving primal problem via dual<br />

The beauty of Corollary 4.2.3.0.1 is its conjugacy; id est, one can solve<br />

either the primal or dual problem in (584) and then find a solution to<br />

the other via the optimality conditions. When a dual optimal solution is<br />

known, for example, a primal optimal solution belongs to the hyperplane<br />

{X | 〈S ⋆ , X〉=0}.<br />

4.2.3.1.1 Example. Minimum cardinality Boolean. [81] [31,4.3.4] [300]<br />

(confer Example 4.5.1.2.1) Consider finding a minimum cardinality Boolean<br />

solution x to the classic linear algebra problem Ax = b given noiseless data<br />

A∈ R m×n and b∈ R m ;<br />

minimize ‖x‖ 0<br />

x<br />

subject to Ax = b<br />

x i ∈ {0, 1} ,<br />

i=1... n<br />

(614)<br />

where ‖x‖ 0 denotes cardinality of vector x (a.k.a, 0-norm; not a convex<br />

function).<br />

A minimum cardinality solution answers the question: “Which fewest<br />

linear combination of columns in A constructs vector b ?” Cardinality<br />

problems have extraordinarily wide appeal, arising in many fields of science<br />

and across many disciplines. [276] [185] [144] [145] Yet designing an efficient<br />

algorithm to optimize cardinality has proved difficult. In this example, we<br />

also constrain the variable to be Boolean. The Boolean constraint forces<br />

4.11 distinct from maximal complementarity (4.1.1.1).

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

Saved successfully!

Ooh no, something went wrong!