Cheng, Tin Sum Lucchi, Aurelien Dokmanić, Ivan Kratsios, Anastasis Belius, David 2023 Existing statistical learning guarantees for general kernel regressors often yield loose bounds when used with finite-rank kernels. Yet, finite-rank kernels naturally appear in several machine learning problems, e.g.\ when fine-tuning a pre-trained deep neural network's last layer to adapt it to a novel task when performing transfer learning. We address this gap for finite-rank kernel ridge regression (KRR) by deriving sharp non-asymptotic upper and lower bounds for the KRR test error of any finite-rank KRR. Our bounds are tighter than previously derived bounds on finite-rank KRR, and unlike comparable results, they also remain valid for any regularization parameters. application/pdf https://n2t.net/ark:/51647/srd1336033 https://sonar.unidistance.ch/global/documents/336033 https://sonar.unidistance.ch/documents/336033/files/A-Theoretical-Analysis-of-the-Test-Error-of-Finite-Rank-Kernel-Ridge-Regression.pdf info:eu-repo/semantics/altIdentifier/doi/10.48550/ARXIV.2310.00987 info:eu-repo/semantics/altIdentifier/ark/51647/srd1336033 info:eu-repo/semantics/openAccess Advances in Neural Information Processing Systems (NeurIPS) 2023. - 2023 A Theoretical Analysis of the Test Error of Finite-Rank Kernel Ridge Regression http://purl.org/coar/resource_type/c_5794