310 followers
@deepcohen It should behave more similar to early stopping in kernel regression. This is of course similar to ridge regularization, but can lead to better convergence rates if the target function is smoother than the kernel: https://t.co/AfGFASXIRI