Efficient SVM training using low-rank kernel representations (2002)

by S Fine, K Scheinberg
Venue:JMLR, 2:243 – 264