We consider the accuracy of an approximate posterior distribution in nonparametric regression problems by combining posterior distributions computed on subsets of the data defined by the locations of the independent variables. We show that this approximate posterior retains the rate of recovery of the full data posterior distribution, where the rate of recovery adapts to the smoothness of the true regression function. As particular examples we consider Gaussian process priors based on integrated Brownian motion and the Mat´ern kernel augmented with a prior on the length scale. Besides theoretical guarantees we present a numerical study of the methods both on synthetic and real world data. We also propose a new aggregation technique, which numerically outperforms previous approaches. Finally, we demonstrate empirically that spatially distributed methods can adapt to local regularities, potentially outperforming the original Gaussian process

Adaptation using spatially distributed Gaussian Processes

Szabo, Botond;
In corso di stampa

Abstract

We consider the accuracy of an approximate posterior distribution in nonparametric regression problems by combining posterior distributions computed on subsets of the data defined by the locations of the independent variables. We show that this approximate posterior retains the rate of recovery of the full data posterior distribution, where the rate of recovery adapts to the smoothness of the true regression function. As particular examples we consider Gaussian process priors based on integrated Brownian motion and the Mat´ern kernel augmented with a prior on the length scale. Besides theoretical guarantees we present a numerical study of the methods both on synthetic and real world data. We also propose a new aggregation technique, which numerically outperforms previous approaches. Finally, we demonstrate empirically that spatially distributed methods can adapt to local regularities, potentially outperforming the original Gaussian process
In corso di stampa
Szabo, Botond; Hadji, Amine; van der Vaart, Aad
File in questo prodotto:
File Dimensione Formato  
distributed_final.pdf

non disponibili

Descrizione: paper
Tipologia: Documento in Pre-print (Pre-print document)
Licenza: Copyright dell'editore
Dimensione 858.81 kB
Formato Adobe PDF
858.81 kB Adobe PDF   Visualizza/Apri
Mail - Botond Tibor Szabo - Outlook.pdf

non disponibili

Descrizione: acceptance letter
Tipologia: Allegato per valutazione Bocconi (Attachment for Bocconi evaluation)
Licenza: NON PUBBLICO - Accesso privato/ristretto
Dimensione 80.52 kB
Formato Adobe PDF
80.52 kB Adobe PDF   Visualizza/Apri
distributed_suppl_final.pdf

non disponibili

Descrizione: supplementary material
Tipologia: Documento in Pre-print (Pre-print document)
Licenza: Copyright dell'editore
Dimensione 1.09 MB
Formato Adobe PDF
1.09 MB Adobe PDF   Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11565/4073276
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact