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.

572 APPENDIX A. LINEAR ALGEBRA<br />

The right-eigenvectors of a diagonalizable matrix A∈ R m×m are linearly<br />

independent if and only if the left-eigenvectors are. So, matrix A has<br />

a representation in terms of its right- and left-eigenvectors; from the<br />

diagonalization (1438), assuming 0 eigenvalues are ordered last,<br />

A =<br />

m∑<br />

λ i s i wi T =<br />

i=1<br />

k∑<br />

≤ m<br />

i=1<br />

λ i ≠0<br />

λ i s i w T i (1472)<br />

From the linearly independent dyads theorem (B.1.1.0.2), the dyads {s i w T i }<br />

must be independent because each set of eigenvectors are; hence rankA = k ,<br />

the number of nonzero eigenvalues. Complex eigenvectors and eigenvalues<br />

are common for real matrices, and must come in complex conjugate pairs for<br />

the summation to remain real. Assume that conjugate pairs of eigenvalues<br />

appear in sequence. Given any particular conjugate pair from (1472), we get<br />

the partial summation<br />

λ i s i w T i + λ ∗ i s ∗ iw ∗T<br />

i = 2Re(λ i s i w T i )<br />

= 2 ( Re s i Re(λ i w T i ) − Im s i Im(λ i w T i ) ) (1473)<br />

where A.18 λ ∗ i = λ i+1 , s ∗ i<br />

equivalently written<br />

A = 2 ∑ i<br />

λ ∈ C<br />

λ i ≠0<br />

∆<br />

∆<br />

= s i+1 , and w ∗ i<br />

Re s 2i Re(λ 2i w T 2i) − Im s 2i Im(λ 2i w T 2i) + ∑ j<br />

λ ∈ R<br />

λ j ≠0<br />

∆<br />

= w i+1 . Then (1472) is<br />

λ j s j w T j (1474)<br />

The summation (1474) shows: A is a linear combination of real and imaginary<br />

parts of its right-eigenvectors corresponding to nonzero eigenvalues. The<br />

k vectors {Re s i ∈ R m , Im s i ∈ R m | λ i ≠0, i∈{1... m}} must therefore span<br />

the range of diagonalizable matrix A .<br />

The argument is similar regarding the span of the left-eigenvectors. <br />

A.7.4<br />

0 trace and matrix product<br />

For X,A∈ R M×N<br />

+ (32)<br />

tr(X T A) = 0 ⇔ X ◦ A = A ◦ X = 0 (1475)<br />

A.18 The complex conjugate of w is denoted w ∗ , while its conjugate transpose is denoted<br />

by w H = w ∗T .

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

Saved successfully!

Ooh no, something went wrong!