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.

A.6. SINGULAR VALUE DECOMPOSITION, SVD 567<br />

zero rows or n−η zero columns; the nonincreasingly ordered (possibly 0)<br />

singular values appear along its main diagonal as for compact SVD (1453).<br />

An important geometrical interpretation of SVD is given in Figure 131<br />

for m = n = 2 : The image of the unit sphere under any m × n matrix<br />

multiplication is an ellipse. Considering the three factors of the SVD<br />

separately, note that Q T is a pure rotation of the circle. Figure 131 shows<br />

how the axes q 1 and q 2 are first rotated by Q T to coincide with the coordinate<br />

axes. Second, the circle is stretched by Σ in the directions of the coordinate<br />

axes to form an ellipse. The third step rotates the ellipse by U into its<br />

final position. Note how q 1 and q 2 are rotated to end up as u 1 and u 2 , the<br />

principal axes of the final ellipse. A direct calculation shows that Aq j = σ j u j .<br />

Thus q j is first rotated to coincide with the j th coordinate axis, stretched by<br />

a factor σ j , and then rotated to point in the direction of u j . All of this<br />

is beautifully illustrated for 2 ×2 matrices by the Matlab code eigshow.m<br />

(see [290]).<br />

A direct consequence of the geometric interpretation is that the largest<br />

singular value σ 1 measures the “magnitude” of A (its 2-norm):<br />

‖A‖ 2 = sup ‖Ax‖ 2 = σ 1 (1461)<br />

‖x‖ 2 =1<br />

This means that ‖A‖ 2 is the length of the longest principal semiaxis of the<br />

ellipse.<br />

Expressions for U , Q , and Σ follow readily from (1460),<br />

AA T U = UΣΣ T and A T AQ = QΣ T Σ (1462)<br />

demonstrating that the columns of U are the eigenvectors of AA T and the<br />

columns of Q are the eigenvectors of A T A. −Neil Muller et alii [234]<br />

A.6.4<br />

Pseudoinverse by SVD<br />

Matrix pseudoinverse (E) is nearly synonymous with singular value<br />

decomposition because of the elegant expression, given A = UΣQ T<br />

A † = QΣ †T U T ∈ R n×m (1463)<br />

that applies to all three flavors of SVD, where Σ †<br />

entries of matrix Σ .<br />

simply inverts nonzero

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

Saved successfully!

Ooh no, something went wrong!