An elementary analysis of ridge regression with random design. (arXiv:2203.08564v2 [math.ST] UPDATED)

In this note, we provide an elementary analysis of the prediction error of
ridge regression with random design. The proof is short and self-contained. In
particular, it bypasses the use of Rudelson's deviation inequality for
covariance matrices, through a combination of exchangeability arguments, matrix
perturbation and operator convexity.

analysis arxiv design elementary math random regression ridge