$σ$-Ridge: group regularized ridge regression via empirical Bayes noise level cross-validation

Features in predictive models are not exchangeable, yet common supervised models treat them as such. Here we study ridge regression when the analyst can partition the features into $K$ groups based on external side-information. For example, in high-throughput biology, features may represent gene expression, protein abundance or clinical data and so each feature group represents a distinct modality. The analyst's goal is to choose optimal regularization parameters $\lambda = (\lambda_1, \dotsc, \lambda_K)$ -- one for each group. In this work, we study the impact of $\lambda$ on the predictive risk of group-regularized ridge regression by deriving limiting risk formulae under a high-dimensional random effects model with $p\asymp n$ as $n \to \infty$. Furthermore, we propose a data-driven method for choosing $\lambda$ that attains the optimal asymptotic risk: The key idea is to interpret the residual noise variance $\sigma^2$, as a regularization parameter to be chosen through cross-validation. An empirical Bayes construction maps the one-dimensional parameter $\sigma$ to the $K$-dimensional vector of regularization parameters, i.e., $\sigma \mapsto \widehat{\lambda}(\sigma)$. Beyond its theoretical optimality, the proposed method is practical and runs as fast as cross-validated ridge regression without feature groups ($K=1$).