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.

386 CHAPTER 5. EUCLIDEAN DISTANCE MATRIX<br />

To maintain relative quadrature between points (Figure 104(a)) and to<br />

prevent reflection, it is sufficient that all interpoint distances be specified and<br />

that adjacencies Y (:, 1:2) , Y (:, 2:3) , and Y (:, 3:4) be proportional to 2×2<br />

rotation matrices; any clockwise rotation would ascribe a reflection matrix<br />

characteristic. Counterclockwise rotation is thereby enforced by constraining<br />

equality among diagonal and antidiagonal entries as prescribed by (886);<br />

[ ] 0 1<br />

Y (:,1:3) = Y (:, 2:4) (890)<br />

−1 0<br />

Quadrature quantization of rotation can be regarded as a constraint<br />

on tilt of the smallest Cartesian square containing the diamond as in<br />

Figure 104(c). Our scheme to quantize rotation requires that all square<br />

vertices be described by vectors whose entries are nonnegative when the<br />

square is translated anywhere interior to the nonnegative orthant. We<br />

capture the four square vertices as columns of a product Y C where<br />

C =<br />

⎡<br />

⎢<br />

⎣<br />

1 0 0 1<br />

1 1 0 0<br />

0 1 1 0<br />

0 0 1 1<br />

⎤<br />

⎥<br />

⎦ (891)<br />

Then, assuming a unit-square shroud, the affine constraint<br />

[ ] 1/2<br />

Y C + 1 T ≥ 0 (892)<br />

1/2<br />

quantizes rotation, as desired.<br />

5.5.2.1 Inner-product form invariance<br />

Likewise, D(Θ) (863) is rotation/reflection invariant;<br />

so (884) and (885) similarly apply.<br />

5.5.3 Invariance conclusion<br />

D(Q p Θ) = D(QΘ) = D(Θ) (893)<br />

In the making of an EDM, absolute rotation, reflection, and translation<br />

information is lost. Given an EDM, reconstruction of point position (5.12,<br />

the list X) can be guaranteed correct only in affine dimension r and relative<br />

position. Given a noiseless complete EDM, this isometric reconstruction is<br />

unique in so far as every realization of a corresponding list X is congruent:

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

Saved successfully!

Ooh no, something went wrong!