Quadrature-based features for kernel approximation

Marina Munkhoeva1Yermek Kapushev1Evgeny Burnaev1Ivan Oseledets1, 2

1Skolkovo Institute of Science and Technology2Institute of Numerical Mathematics of the Russian Academy of Sciences

Conference on Neural Information Processing Systems 2018

Error of the approximation of the kernel functions


We consider the problem of improving kernel approximation via randomized feature maps. These maps arise as Monte Carlo approximation to integral representations of kernel functions and scale up kernel methods for larger datasets. Based on an efficient numerical integration technique, we propose a unifying approach that reinterprets the previous random features methods and extends to better estimates of the kernel approximation. We derive the convergence behaviour and conduct an extensive empirical study that supports our hypothesis.




Copy bibtex