20.05.2014 Views

link to my thesis

link to my thesis

link to my thesis

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.

7.1. COMPUTING THE MINIMUM OF A POLYNOMIAL OF ORDER 8 109<br />

given by:<br />

x 1 =+0.876539<br />

x 2 = −0.903966<br />

x 3 =+0.862028<br />

x 4 = −0.835187<br />

(7.3)<br />

(ii) In an attempt <strong>to</strong> obtain similar information in an independent fashion, the<br />

matrices A xi for all i = 1,...,4, and A p1 have been constructed explicitly using<br />

exact computation. The involved quotient space R[x 1 ,x 2 ,x 3 ,x 4 ]/I is of dimension<br />

(2d − 1) n = 2401. Therefore these matrices have 2401 eigenvalues. A property of<br />

these matrices is that they are highly sparse: the matrix A x1 contains only 4619<br />

non-zero elements (percentage filled: 0.08%), whereas the matrix A p1 contains 43178<br />

non-zero elements (percentage filled: 0.75%). See Figure 7.1 for a representation of<br />

the sparsity structure of the matrices A x1 and A p1 .<br />

Figure 7.1: Sparsity structure of the matrices A x1<br />

and A p1<br />

Building the matrix A p1 <strong>to</strong>ok 465 seconds (whereas building the matrices A x1 ,<br />

A x2 , A x3 and A x4 <strong>to</strong>ok 56, 60, 63, and 61 seconds, respectively). All the eigenvalues<br />

of the matrix A p1 are computed numerically in two ways: using the eigenvalue solver<br />

Eigensystem in Mathematica and the eigenvalue solver Eig in Matlab. In Table 7.1<br />

the results of these computations are collected.<br />

Table 7.1: Minimal real eigenvalue of the explicit matrix A p1<br />

Method Eigenvalue Time (s)<br />

NSolve solution of system (7.2) 4.09516474435915 744<br />

Eigensystem construct A p1 , compute all eigenvalues 4.09516474435915 465 + 315<br />

Eig construct A p1 , compute all eigenvalues 4.09516474435924 465 + 176<br />

The outcomes of these computations agree up <strong>to</strong> a limited number of decimal digits<br />

(the fact that the outcomes of the NSolve and the Eigenvalue function in Mathematica

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

Saved successfully!

Ooh no, something went wrong!