13.07.2015 Views

Mathematical Methods for Physics and Engineering - Matematica.NET

Mathematical Methods for Physics and Engineering - Matematica.NET

Mathematical Methods for Physics and Engineering - Matematica.NET

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.

8.18 SIMULTANEOUS LINEAR EQUATIONSWe already know from the above discussion, however, that the non-zero eigenvalues ofthis matrix are equal to those of AA † found above, <strong>and</strong> that the remaining eigenvalue iszero. The corresponding normalised eigenvectors are easily found:λ 1 =16 ⇒ v 1 = 1 (1 1 1 1)T2<strong>and</strong>sothematrixV is given byλ 2 =9 ⇒ v 2 = 1 (1 1 − 1 − 1)T2λ 3 =4 ⇒ v 3 = 1 (−1 1 1 − 1)T2λ 4 =0 ⇒ v 4 = 1 (1 − 1 1 − 1)T2⎛V = 1 ⎜⎝2⎞1 1 −1 11 1 1 −1 ⎟1 −1 1 1⎠ . (8.140)1 −1 −1 −1Alternatively, we could have found the first three columns of V by using the relation(8.135) to obtainv i = 1 A † u i <strong>for</strong> i =1, 2, 3.s iThe fourth eigenvector could then be found using the Gram–Schmidt orthogonalisationprocedure. We note that if there were more than one eigenvector corresponding to a zeroeigenvalue then we would need to use this procedure to orthogonalise these eigenvectorsbe<strong>for</strong>e constructing the matrix V.Collecting our results together, we find the SVD of the matrix A:⎛1 0 0A = USV † ⎜= ⎝ 03045− 4 5 535⎛⎞ ⎛⎟⎠ ⎝ 4 0 0 0⎞0 3 0 0 ⎠⎜0 0 2 0 ⎝this can be verified by direct multiplication. ◭1212− 1 212112122− 1 2− 1 21212− 1 212− 1 212− 1 2Let us now consider the use of SVD in solving a set of M simultaneous linearequations in N unknowns, which we write again as Ax = b. Firstly, considerthe solution of a homogeneous set of equations, <strong>for</strong> which b = 0. As mentionedpreviously, if A is square <strong>and</strong> non-singular (<strong>and</strong> so possesses no zero singularvalues) then the equations have the unique trivial solution x = 0. Otherwise,anyof the vectors v i , i = r +1,r+2,...,N, or any linear combination of them, willbe a solution.In the inhomogeneous case, where b is not a zero vector, the set of equationswill possess solutions if b lies in the range of A. To investigate these solutions, itis convenient to introduce the N × M matrix S, which is constructed by takingthe transpose of S in (8.131) <strong>and</strong> replacing each non-zero singular value s i onthe diagonal by 1/s i . It is clear that, with this construction, SS is an M × Mdiagonal matrix with diagonal entries that equal unity <strong>for</strong> those values of j <strong>for</strong>which s j ≠ 0, <strong>and</strong> zero otherwise.Now consider the vectorˆx = VSU † b. (8.141)305⎞⎟⎠ ;

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

Saved successfully!

Ooh no, something went wrong!