Automatic Construction of Nonparametric Relational Regression Models for Multiple Time Series

Yunseong Hwang, Anh Tong, Jaesik Choi
Proceedings of The 33rd International Conference on Machine Learning, PMLR 48:3030-3039, 2016.

Abstract

Gaussian Processes (GPs) provide a general and analytically tractable way of modeling complex time-varying, nonparametric functions. The Automatic Bayesian Covariance Discovery (ABCD) system constructs natural-language description of time-series data by treating unknown time-series data nonparametrically using GP with a composite covariance kernel function. Unfortunately, learning a composite covariance kernel with a single time-series data set often results in less informative kernel that may not give qualitative, distinctive descriptions of data. We address this challenge by proposing two relational kernel learning methods which can model multiple time-series data sets by finding common, shared causes of changes. We show that the relational kernel learning methods find more accurate models for regression problems on several real-world data sets; US stock data, US house price index data and currency exchange rate data.

Cite this Paper

BibTeX
@InProceedings{pmlr-v48-hwangb16, title = {Automatic Construction of Nonparametric Relational Regression Models for Multiple Time Series}, author = {Hwang, Yunseong and Tong, Anh and Choi, Jaesik}, booktitle = {Proceedings of The 33rd International Conference on Machine Learning}, pages = {3030--3039}, year = {2016}, editor = {Balcan, Maria Florina and Weinberger, Kilian Q.}, volume = {48}, series = {Proceedings of Machine Learning Research}, address = {New York, New York, USA}, month = {20--22 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v48/hwangb16.pdf}, url = {https://proceedings.mlr.press/v48/hwangb16.html}, abstract = {Gaussian Processes (GPs) provide a general and analytically tractable way of modeling complex time-varying, nonparametric functions. The Automatic Bayesian Covariance Discovery (ABCD) system constructs natural-language description of time-series data by treating unknown time-series data nonparametrically using GP with a composite covariance kernel function. Unfortunately, learning a composite covariance kernel with a single time-series data set often results in less informative kernel that may not give qualitative, distinctive descriptions of data. We address this challenge by proposing two relational kernel learning methods which can model multiple time-series data sets by finding common, shared causes of changes. We show that the relational kernel learning methods find more accurate models for regression problems on several real-world data sets; US stock data, US house price index data and currency exchange rate data.} }
Endnote
%0 Conference Paper %T Automatic Construction of Nonparametric Relational Regression Models for Multiple Time Series %A Yunseong Hwang %A Anh Tong %A Jaesik Choi %B Proceedings of The 33rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2016 %E Maria Florina Balcan %E Kilian Q. Weinberger %F pmlr-v48-hwangb16 %I PMLR %P 3030--3039 %U https://proceedings.mlr.press/v48/hwangb16.html %V 48 %X Gaussian Processes (GPs) provide a general and analytically tractable way of modeling complex time-varying, nonparametric functions. The Automatic Bayesian Covariance Discovery (ABCD) system constructs natural-language description of time-series data by treating unknown time-series data nonparametrically using GP with a composite covariance kernel function. Unfortunately, learning a composite covariance kernel with a single time-series data set often results in less informative kernel that may not give qualitative, distinctive descriptions of data. We address this challenge by proposing two relational kernel learning methods which can model multiple time-series data sets by finding common, shared causes of changes. We show that the relational kernel learning methods find more accurate models for regression problems on several real-world data sets; US stock data, US house price index data and currency exchange rate data.
RIS
TY - CPAPER TI - Automatic Construction of Nonparametric Relational Regression Models for Multiple Time Series AU - Yunseong Hwang AU - Anh Tong AU - Jaesik Choi BT - Proceedings of The 33rd International Conference on Machine Learning DA - 2016/06/11 ED - Maria Florina Balcan ED - Kilian Q. Weinberger ID - pmlr-v48-hwangb16 PB - PMLR DP - Proceedings of Machine Learning Research VL - 48 SP - 3030 EP - 3039 L1 - http://proceedings.mlr.press/v48/hwangb16.pdf UR - https://proceedings.mlr.press/v48/hwangb16.html AB - Gaussian Processes (GPs) provide a general and analytically tractable way of modeling complex time-varying, nonparametric functions. The Automatic Bayesian Covariance Discovery (ABCD) system constructs natural-language description of time-series data by treating unknown time-series data nonparametrically using GP with a composite covariance kernel function. Unfortunately, learning a composite covariance kernel with a single time-series data set often results in less informative kernel that may not give qualitative, distinctive descriptions of data. We address this challenge by proposing two relational kernel learning methods which can model multiple time-series data sets by finding common, shared causes of changes. We show that the relational kernel learning methods find more accurate models for regression problems on several real-world data sets; US stock data, US house price index data and currency exchange rate data. ER -
APA
Hwang, Y., Tong, A. & Choi, J.. (2016). Automatic Construction of Nonparametric Relational Regression Models for Multiple Time Series. Proceedings of The 33rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 48:3030-3039 Available from https://proceedings.mlr.press/v48/hwangb16.html.

Related Material