Ridge Ensemble

Ridge ensemble methods combine predictions from multiple ridge regression models trained on different subsets of features, aiming to improve prediction accuracy and stability. Current research focuses on establishing theoretical equivalences between feature subsampling and ridge regularization within these ensembles, particularly analyzing the impact of subsampling ratios and regularization parameters on prediction risk. These studies leverage proportional asymptotics and generalized cross-validation to optimize ensemble performance and demonstrate connections between optimal ridge regression and optimally tuned ridgeless ensembles. This work contributes to a deeper understanding of high-dimensional regression and provides practical tools for improving the efficiency and accuracy of prediction models.