10.07.2015 Views

v2007.09.13 - Convex Optimization

v2007.09.13 - Convex Optimization

v2007.09.13 - 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.

6.8. DUAL EDM CONE 425whose veracity is intuitively evident, in hindsight, [61, p.109] from themost fundamental EDM definition (705). Formula (1070) is not a matrixcriterion for membership to the EDM cone, and it is not an equivalencebetween EDM operators. Rather, it is a recipe for constructing the EDMcone whole from large Euclidean bodies: the positive semidefinite cone,orthogonal complement of the geometric center subspace, and symmetrichollow subspace. A realization of this construction in low dimension isillustrated in Figure 104 and Figure 105.The dual EDM cone follows directly from (1070) by standard propertiesof cones (2.13.1.1):EDM N∗ = K ∗ 1 + K ∗ 2 = S N⊥h − S N c ∩ S N + (1071)which bears strong resemblance to (1050).6.8.1.2 nonnegative orthantThat EDM N is a proper subset of the nonnegative orthant is not obviousfrom (1070). We wish to verifyEDM N = S N h ∩ ( )S N⊥c − S N + ⊂ RN×N+ (1072)While there are many ways to prove this, it is sufficient to show that allentries of the extreme directions of EDM N must be nonnegative; id est, forany particular nonzero vector z = [z i , i=1... N]∈ N(1 T ) (6.5.3.1),δ(zz T )1 T + 1δ(zz T ) T − 2zz T ≥ 0 (1073)where the inequality denotes entrywise comparison. The inequality holdsbecause the i,j th entry of an extreme direction is squared: (z i − z j ) 2 .We observe that the dyad 2zz T ∈ S N + belongs to the positive semidefinitecone, the doubletδ(zz T )1 T + 1δ(zz T ) T ∈ S N⊥c (1074)to the orthogonal complement (1763) of the geometric center subspace, whiletheir difference is a member of the symmetric hollow subspace S N h .Here is an algebraic method provided by Trosset to prove nonnegativity:Suppose we are given A∈ S N⊥c and B = [B ij ]∈ S N + and A −B ∈ S N h .Then we have, for some vector u , A = u1 T + 1u T = [A ij ] = [u i + u j ] andδ(B)= δ(A)= 2u . Positive semidefiniteness of B requires nonnegativityA −B ≥ 0 because(e i −e j ) T B(e i −e j ) = (B ii −B ij )−(B ji −B jj ) = 2(u i +u j )−2B ij ≥ 0 (1075)

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

Saved successfully!

Ooh no, something went wrong!