Time Series Prediction Based on the Relevance Vector Machine with Adaptive Kernels

Joaquin Quinonero; Lars Kai Hansen

doi:10.1109/icassp.2002.5743959

Time Series Prediction Based on the Relevance Vector Machine with Adaptive Kernels

Joaquin Quinonero, Lars Kai Hansen

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

429 Downloads (Orbit)

Abstract

The Relevance Vector Machine (RVM) introduced by Tipping is a probabilistic model similar to the widespread Support Vector Machines (SVM), but where the training takes place in a Bayesian framework, and where predictive distributions of the outputs instead of point estimates are obtained. In this paper we focus on the use of RVM's for regression. We modify this method for training generalized linear models by adapting automatically the width of the basis functions to the optimal for the data at hand. Our Adaptive RVM is tried for prediction on the chaotic Mackey-Glass time series. Much superior performance than with the standard RVM and than with other methods like neural networks and local linear models is obtained.

Original language	English
Title of host publication	International Conference on Acoustics, Speech, and Signal Processing
Publication date	2002
Pages	985-988
DOIs	https://doi.org/10.1109/icassp.2002.5743959
Publication status	Published - 2002
Event	2002 IEEE International Conference on Acoustics, Speech, and Signal Processing - Orlando, United States Duration: 13 May 2002 → 17 May 2002 Conference number: 27

Conference

Conference	2002 IEEE International Conference on Acoustics, Speech, and Signal Processing
Number	27
Country/Territory	United States
City	Orlando
Period	13/05/2002 → 17/05/2002

Access to Document

10.1109/icassp.2002.5743959

FulltextSubmitted manuscript, 156 KB
FulltextSubmitted manuscript, 157 KB

OpenUrl availability

Full text

Cite this

@inproceedings{24bde21e493d4bd7aed3a8d0f2780a08,

title = "Time Series Prediction Based on the Relevance Vector Machine with Adaptive Kernels",

abstract = "The Relevance Vector Machine (RVM) introduced by Tipping is a probabilistic model similar to the widespread Support Vector Machines (SVM), but where the training takes place in a Bayesian framework, and where predictive distributions of the outputs instead of point estimates are obtained. In this paper we focus on the use of RVM's for regression. We modify this method for training generalized linear models by adapting automatically the width of the basis functions to the optimal for the data at hand. Our Adaptive RVM is tried for prediction on the chaotic Mackey-Glass time series. Much superior performance than with the standard RVM and than with other methods like neural networks and local linear models is obtained.",

keywords = "Kernel Methods, adaptive kernels, Bayesian Methods, relevance vector machine, Time series prediction",

author = "Joaquin Quinonero and Hansen, \{Lars Kai\}",

year = "2002",

doi = "10.1109/icassp.2002.5743959",

language = "English",

pages = "985--988",

booktitle = "International Conference on Acoustics, Speech, and Signal Processing",

note = "2002 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP '02 ; Conference date: 13-05-2002 Through 17-05-2002",

}

Quinonero, J & Hansen, LK 2002, Time Series Prediction Based on the Relevance Vector Machine with Adaptive Kernels. in International Conference on Acoustics, Speech, and Signal Processing. pp. 985-988, 2002 IEEE International Conference on Acoustics, Speech, and Signal Processing, Orlando, Florida, United States, 13/05/2002. https://doi.org/10.1109/icassp.2002.5743959

TY - GEN

T1 - Time Series Prediction Based on the Relevance Vector Machine with Adaptive Kernels

AU - Quinonero, Joaquin

AU - Hansen, Lars Kai

N1 - Conference code: 27

PY - 2002

Y1 - 2002

N2 - The Relevance Vector Machine (RVM) introduced by Tipping is a probabilistic model similar to the widespread Support Vector Machines (SVM), but where the training takes place in a Bayesian framework, and where predictive distributions of the outputs instead of point estimates are obtained. In this paper we focus on the use of RVM's for regression. We modify this method for training generalized linear models by adapting automatically the width of the basis functions to the optimal for the data at hand. Our Adaptive RVM is tried for prediction on the chaotic Mackey-Glass time series. Much superior performance than with the standard RVM and than with other methods like neural networks and local linear models is obtained.

AB - The Relevance Vector Machine (RVM) introduced by Tipping is a probabilistic model similar to the widespread Support Vector Machines (SVM), but where the training takes place in a Bayesian framework, and where predictive distributions of the outputs instead of point estimates are obtained. In this paper we focus on the use of RVM's for regression. We modify this method for training generalized linear models by adapting automatically the width of the basis functions to the optimal for the data at hand. Our Adaptive RVM is tried for prediction on the chaotic Mackey-Glass time series. Much superior performance than with the standard RVM and than with other methods like neural networks and local linear models is obtained.

KW - Kernel Methods

KW - adaptive kernels

KW - Bayesian Methods

KW - relevance vector machine

KW - Time series prediction

U2 - 10.1109/icassp.2002.5743959

DO - 10.1109/icassp.2002.5743959

M3 - Article in proceedings

SP - 985

EP - 988

BT - International Conference on Acoustics, Speech, and Signal Processing

T2 - 2002 IEEE International Conference on Acoustics, Speech, and Signal Processing

Y2 - 13 May 2002 through 17 May 2002

ER -

Time Series Prediction Based on the Relevance Vector Machine with Adaptive Kernels

Abstract

Conference

Access to Document

OpenUrl availability

Fingerprint

Cite this