COLT
2001
Springer
14 years 4 months ago
2001
Springer
Abstract. We develop three new techniques to build on the recent advances in online learning with kernels. First, we show that an exponential speed-up in prediction time per trial ...