Tech Report on Regularized Least Squares
An MIT tech report with things everyone ought to know about (dense) regularized least squares is available in the publications section. It includes all the goodies on how finding a good a regularization parameter via leave-one-out cross-validation is essentially free, with a couple of different proofs.
I’m not trying to have this published at the conference or journal level because I don’t think it contains (enough) “new” material, but I certainly haven’t seen these results usefully collected together like this before. I hope people find this useful.
May 5, 01:03 pm | machine-learning |
Comments
commenting closed for this article