Singular Value Decomposition And Least Squares Solutions (With C. Reinsch)
Raymond H. Chan, Chen Greif, Dianne Prost O’Leary · 2007
Abstract method, then state of the art. We see that the singular values are the square roots of the eigenvalues of the positive (semi)definite matrices AHA or A AH. This means numerical difficulties with the accurate computation of the smaller singular values, which usually are the more interesting ones and should be computed as precisely as possible. An equally severe problem is the correct association of the eigenvectors of AHA (columns of V ) with the eigenvectors of A AH (columns of U ) which are not unique in the event of multiple eigenvalues.