Anytime Interval-Valued Outputs for Kernel Machines: Fast Support Vector Machine Classification via Distance Geometry

Classifying M examples using a support vector machine containing L support vectors traditionally requires exactly M . L kernel computations. We introduce a computational geometry method for which classification cost becomes roughly proportional to the difficulty of each example (e.g. distance from the discriminant hyperplane). It produces exactly the same classifications, while typically requiring much (e.g. 10 times) fewer kernel computations than Ma L. Related 5educed set” methods (e.g. (Burges, 1996; Scholkopf et al., 1999; Scholkopf et al., 1998)) similarly lower the effective L, but provide neither proportionality with difficulty nor guaranteed preservation of classifications.