Samy Bengio, Alexei Pozdnoukhov
This paper presents a general method for incorporating prior knowledge into kernel methods. It applies when the prior knowledge can be formalized by the description of an object around each sample of the training set, assuming that all points in the given ...
IDIAP2003