NIPS
2007
13 years 6 months ago
2007
To accelerate the training of kernel machines, we propose to map the input data to a randomized low-dimensional feature space and then apply existing fast linear methods. The feat...