On the Optimality of Random Features under General Conditions: Misspecified Cases and Robust Learning

Caixing Wang

Random features provide a standard approach to scaling kernel methods to large datasets by approximating the kernel with a finite-dimensional feature map. Existing works of random…