One should not feel too happy with the presented alternative
algorithm(s) for computing the singular value decomposition (SVD) of
normal matrices. Although the key idea is interesting (viz., trying
to make an explicit use of the normality and introducing a detour
via complex symmetric matrices) and although its implementation
leads to some interesting experimental observations (viz., an
amendment of the memory consumption and a certain lowering of the
computational complexity), the key weakness of the whole approach
(viz., a sensitivity to the clustered eigenvalues and an apparent
loss of competitiveness in accuracy) is just reconfirmed by the
authors conclusion that ``further research is required''.

MR3252840 Ferranti, Micol; Le, Thanh Hieu; Vandebril, Raf A
comparison between the complex symmetric based and classical
computation of the singular value decomposition of normal matrices.
Numer. Algorithms 67 (2014), no. 1, 109--120. 65F15