Date(s) : 14/02/2014 iCal
14 h 00 min - 15 h 00 min
Operator-valued kernels, such as multi-task kernels, are appropriate for learning problems with nonscalar outputs like structured output prediction, and have interesting properties. For instance, they are particularly efficient for dealing with functional data, and for taking into account the structure of the output space. However, they have received little attention from the community, particularly from a theoretical point of vue. We will show that operator-valued kernel regression algorithms are uniformly stable in the general case of infinite-dimensional output spaces. We then derive under mild assumption on the kernel generalization bounds of such algorithms, and we show their consistency even with non Hilbert-Schmidt operator-valued kernels.
http://pageperso.lif.univ-mrs.fr/~julien.audiffren/« >http://pageperso.lif.univ-mrs.fr/~julien.audiffren/
Catégories Pas de Catégories