Finite-dimensional approximation of Gaussian processes

Giancarlo Ferrari Trecate; Christopher K.I. Williams; Manfred Opper

Finite-dimensional approximation of Gaussian processes

Giancarlo Ferrari Trecate, Christopher K.I. Williams, Manfred Opper

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract / Description of output

Gaussian process (GP) prediction suffers from O(n3) scaling with thedata set size n. By using a finite-dimensional basis to approximate theGP predictor, the computational complexity can be reduced. We deriveoptimal finite-dimensional predictors under a number of assumptions,and show the superiority of these predictors over the ProjectedBayes Regression method (which is asymptotically optimal). We alsoshow how to calculate the minimal model size for a given n. Thecalculations are backed up by numerical experiments.

Original language	English
Title of host publication	Advances in Neural Information Processing Systems 11 (NIPS 1998)
Publisher	MIT Press
Pages	218-224
Number of pages	7
Publication status	Published - 1999

Access to Document

http://papers.nips.cc/paper/1497-finite-dimensional-approximation-of-gaussian-processes

Cite this

@inproceedings{197fb5fe43e3471bb128362f419dfe0e,

title = "Finite-dimensional approximation of Gaussian processes",

abstract = "Gaussian process (GP) prediction suffers from O(n3) scaling with thedata set size n. By using a finite-dimensional basis to approximate theGP predictor, the computational complexity can be reduced. We deriveoptimal finite-dimensional predictors under a number of assumptions,and show the superiority of these predictors over the ProjectedBayes Regression method (which is asymptotically optimal). We alsoshow how to calculate the minimal model size for a given n. Thecalculations are backed up by numerical experiments.",

author = "Trecate, {Giancarlo Ferrari} and Williams, {Christopher K.I.} and Manfred Opper",

year = "1999",

language = "English",

pages = "218--224",

booktitle = "Advances in Neural Information Processing Systems 11 (NIPS 1998)",

publisher = "MIT Press",

}

TY - GEN

T1 - Finite-dimensional approximation of Gaussian processes

AU - Trecate, Giancarlo Ferrari

AU - Williams, Christopher K.I.

AU - Opper, Manfred

PY - 1999

Y1 - 1999

N2 - Gaussian process (GP) prediction suffers from O(n3) scaling with thedata set size n. By using a finite-dimensional basis to approximate theGP predictor, the computational complexity can be reduced. We deriveoptimal finite-dimensional predictors under a number of assumptions,and show the superiority of these predictors over the ProjectedBayes Regression method (which is asymptotically optimal). We alsoshow how to calculate the minimal model size for a given n. Thecalculations are backed up by numerical experiments.

AB - Gaussian process (GP) prediction suffers from O(n3) scaling with thedata set size n. By using a finite-dimensional basis to approximate theGP predictor, the computational complexity can be reduced. We deriveoptimal finite-dimensional predictors under a number of assumptions,and show the superiority of these predictors over the ProjectedBayes Regression method (which is asymptotically optimal). We alsoshow how to calculate the minimal model size for a given n. Thecalculations are backed up by numerical experiments.

M3 - Conference contribution

SP - 218

EP - 224

BT - Advances in Neural Information Processing Systems 11 (NIPS 1998)

PB - MIT Press

ER -

Finite-dimensional approximation of Gaussian processes

Abstract / Description of output

Access to Document

Fingerprint

Cite this