JUST INTERPOLATE: KERNEL “RIDGELESS” REGRESSION CAN GENERALIZE

In the absence of explicit regularization, Kernel “Ridgeless” Regression with nonlinear kernels has the potential to fit the training data perfectly. It has been observed empirically, however, that such interpolated solutions can still generalize well on test data. We isolate a phenomenon of implici...

Ausführliche Beschreibung

Bibliographische Detailangaben

Veröffentlicht in:	The Annals of statistics 2020-06, Vol.48 (3), p.1329-1347
Hauptverfasser:	Liang, Tengyuan, Rakhlin, Alexander
Format:	Artikel
Sprache:	eng
Schlagworte:	Asymptotic methods Covariance Datasets Eigenvalues Kernel functions Mathematical functions Nonlinear systems Regression analysis Regularization Sampling error Upper bounds
Online-Zugang:	Volltext
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!