Connectionists: matrix SVD and kernel methods

[Johan Suykens]

Dear all,


I would like to announce the following publication:


Suykens J.A.K., "SVD revisited: a new variational principle, compatible 
feature maps and nonlinear extensions", Applied and Computational 
Harmonic Analysis, 40 (2016), pp. 600-609. doi:10.1016/j.acha.2015.09.004
http://www.sciencedirect.com/science/article/pii/S1063520315001360
http://www.esat.kuleuven.be/stadius/ADB/publications.php


- it proposes a new variational principle for the matrix singular
value decomposition, in a kernel-based learning setting;

- it goes beyond Mercer kernels (which are commonly used in the
kernel trick);

- nonlinear extensions are shown to the matrix SVD

- kernel PCA corresponds to a special case of the formulations, related 
to the case of a square symmetric matrix.


Best regards,
Johan Suykens


----------------------

Prof. Johan Suykens
Katholieke Universiteit Leuven
Departement Elektrotechniek - ESAT-STADIUS
Kasteelpark Arenberg 10
B-3001 Leuven (Heverlee)
Belgium
http://www.esat.kuleuven.be/stadius/members/suykens.html
http://www.esat.kuleuven.be/stadius/ADB