e-Distance Weighted Support Vector Regression

Yan Wang; Ge Ou; Wei Pang; Lan Huang; George MacLeod Coghill

e-Distance Weighted Support Vector Regression

Yan Wang, Ge Ou, Wei Pang, Lan Huang, George MacLeod Coghill

Jilin University

Research output: Working paper

Abstract

We propose a novel support vector regression approach called e-Distance Weighted Support Vector Regression (e-DWSVR).e-DWSVR specifically addresses two challenging issues in support vector regression: first, the process of noisy data; second, how to deal with the situation when the distribution of boundary data is different from that of the overall data. The proposed e-DWSVR optimizes the minimum margin and the mean of functional margin simultaneously to tackle these two issues. In addition, we use both dual coordinate descent (CD) and averaged stochastic gradient descent (ASGD) strategies to make e-DWSVR scalable to large scale problems. We report promising results obtained by e-DWSVR in comparison with existing methods on several benchmark datasets.

Original language	English
Publisher	ArXiv
Pages	1-10
Number of pages	10
Publication status	Published - 21 Jul 2016

Bibliographical note

We gratefully thank Dr Teng Zhang and Prof Zhi-Hua Zhou for providing the source code of “LDM” source code and their kind technical assistance. We also thank Prof Chih-Jen Lins team for providing the Libsvm and Liblinear packages and their support. This work is supported by the National Natural Science Foundation of China (Nos. 61472159, 61572227) and Development Project of Jilin Province of China (Nos.20140101180JC, 20160204022GX). This work is also partially supported by the 2015 Scottish Crucible Award funded by the Royal Society of Edinburgh and the 2016 PECE bursary provided by the Scottish Informatics & Computer Science Alliance (SICSA).

Access to Document

https://arxiv.org/abs/1607.06657Licence: Unspecified

Cite this

@techreport{2f397a3b9c504e39ad18f37f752670e9,

title = "e-Distance Weighted Support Vector Regression",

abstract = "We propose a novel support vector regression approach called e-Distance Weighted Support Vector Regression (e-DWSVR).e-DWSVR specifically addresses two challenging issues in support vector regression: first, the process of noisy data; second, how to deal with the situation when the distribution of boundary data is different from that of the overall data. The proposed e-DWSVR optimizes the minimum margin and the mean of functional margin simultaneously to tackle these two issues. In addition, we use both dual coordinate descent (CD) and averaged stochastic gradient descent (ASGD) strategies to make e-DWSVR scalable to large scale problems. We report promising results obtained by e-DWSVR in comparison with existing methods on several benchmark datasets.",

author = "Yan Wang and Ge Ou and Wei Pang and Lan Huang and Coghill, {George MacLeod}",

note = "We gratefully thank Dr Teng Zhang and Prof Zhi-Hua Zhou for providing the source code of “LDM” source code and their kind technical assistance. We also thank Prof Chih-Jen Lins team for providing the Libsvm and Liblinear packages and their support. This work is supported by the National Natural Science Foundation of China (Nos. 61472159, 61572227) and Development Project of Jilin Province of China (Nos.20140101180JC, 20160204022GX). This work is also partially supported by the 2015 Scottish Crucible Award funded by the Royal Society of Edinburgh and the 2016 PECE bursary provided by the Scottish Informatics & Computer Science Alliance (SICSA).",

year = "2016",

month = jul,

day = "21",

language = "English",

pages = "1--10",

publisher = "ArXiv",

type = "WorkingPaper",

institution = "ArXiv",

}

TY - UNPB

T1 - e-Distance Weighted Support Vector Regression

AU - Wang, Yan

AU - Ou, Ge

AU - Pang, Wei

AU - Huang, Lan

AU - Coghill, George MacLeod

N1 - We gratefully thank Dr Teng Zhang and Prof Zhi-Hua Zhou for providing the source code of “LDM” source code and their kind technical assistance. We also thank Prof Chih-Jen Lins team for providing the Libsvm and Liblinear packages and their support. This work is supported by the National Natural Science Foundation of China (Nos. 61472159, 61572227) and Development Project of Jilin Province of China (Nos.20140101180JC, 20160204022GX). This work is also partially supported by the 2015 Scottish Crucible Award funded by the Royal Society of Edinburgh and the 2016 PECE bursary provided by the Scottish Informatics & Computer Science Alliance (SICSA).

PY - 2016/7/21

Y1 - 2016/7/21

N2 - We propose a novel support vector regression approach called e-Distance Weighted Support Vector Regression (e-DWSVR).e-DWSVR specifically addresses two challenging issues in support vector regression: first, the process of noisy data; second, how to deal with the situation when the distribution of boundary data is different from that of the overall data. The proposed e-DWSVR optimizes the minimum margin and the mean of functional margin simultaneously to tackle these two issues. In addition, we use both dual coordinate descent (CD) and averaged stochastic gradient descent (ASGD) strategies to make e-DWSVR scalable to large scale problems. We report promising results obtained by e-DWSVR in comparison with existing methods on several benchmark datasets.

AB - We propose a novel support vector regression approach called e-Distance Weighted Support Vector Regression (e-DWSVR).e-DWSVR specifically addresses two challenging issues in support vector regression: first, the process of noisy data; second, how to deal with the situation when the distribution of boundary data is different from that of the overall data. The proposed e-DWSVR optimizes the minimum margin and the mean of functional margin simultaneously to tackle these two issues. In addition, we use both dual coordinate descent (CD) and averaged stochastic gradient descent (ASGD) strategies to make e-DWSVR scalable to large scale problems. We report promising results obtained by e-DWSVR in comparison with existing methods on several benchmark datasets.

M3 - Working paper

SP - 1

EP - 10

BT - e-Distance Weighted Support Vector Regression

PB - ArXiv

ER -

e-Distance Weighted Support Vector Regression

Abstract

Bibliographical note

Access to Document

Fingerprint

Cite this