Random Features for Large-Scale Kernel Machines

← Back to document page

of the data [4], or produce good low-rank or sparse approximations of the true kernel matrix [3, 7]. Fast multipole and multigrid methods have also been proposed for this purpose, but, while they ap-pear to be effective on small and low-dimensional problems, to our knowledge, their effectiveness has not been demonstrated on large datasets.

  Feature, Methods, Arsesp