In this paper, we study regression problems over a separable Hilbert space with the square loss, covering non-parametric regression over a reproducing kernel Hilbert space. We investigate a class of spectral/regularized algorithms, including ridge regression, principal component regression, and gradient methods. We prove optimal, high-probability convergence results in terms of variants of norms for the studied algorithms, considering a capacity assumption on the hypothesis space and a general source condition on the target function. Consequently, we obtain almost sure convergence results with optimal rates. Our results improve and generalize previous results, filling a theoretical gap for the non-attainable cases
Optimal rates for spectral algorithms with least-squares regression over Hilbert spaces
Rosasco, Lorenzo;
2020-01-01
Abstract
In this paper, we study regression problems over a separable Hilbert space with the square loss, covering non-parametric regression over a reproducing kernel Hilbert space. We investigate a class of spectral/regularized algorithms, including ridge regression, principal component regression, and gradient methods. We prove optimal, high-probability convergence results in terms of variants of norms for the studied algorithms, considering a capacity assumption on the hypothesis space and a general source condition on the target function. Consequently, we obtain almost sure convergence results with optimal rates. Our results improve and generalize previous results, filling a theoretical gap for the non-attainable casesFile | Dimensione | Formato | |
---|---|---|---|
1801.06720.pdf
accesso aperto
Tipologia:
Documento in Post-print
Dimensione
319.87 kB
Formato
Adobe PDF
|
319.87 kB | Adobe PDF | Visualizza/Apri |
1-s2.0-S1063520318300174-main.pdf
accesso chiuso
Tipologia:
Documento in versione editoriale
Dimensione
528.98 kB
Formato
Adobe PDF
|
528.98 kB | Adobe PDF | Visualizza/Apri Richiedi una copia |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.