Transfer dimensionality reduction by Gaussian process in parallel

Bin Tong, Junbin Gao, Thach Nguyen Huy, Hao Shao, Einoshin Suzuki

Research output: Contribution to journalArticle

Abstract

Dimensionality reduction has been considered as one of the most significant tools for data analysis. In general, supervised information is helpful for dimensionality reduction. However, in typical real applications, supervised information in multiple source tasks may be available, while the data of the target task are unlabeled. An interesting problem of how to guide the dimensionality reduction for the unlabeled target data by exploiting useful knowledge, such as label information, from multiple source tasks arises in such a scenario. In this paper, we propose a new method for dimensionality reduction in the transfer learning setting. Unlike traditional paradigms where the useful knowledge from multiple source tasks is transferred through distance metric, we attempt to learn a more informative mapping function between the original data and the reduced data by Gaussian process that behaves more appropriately than other parametric regression methods due to its less parametric characteristic. In our proposal, we firstly convert the dimensionality reduction problem into integral regression problems in parallel. Gaussian process is then employed to learn the underlying relationship between the original data and the reduced data. Such a relationship can be appropriately transferred to the target task by exploiting the prediction ability of the Gaussian process model and inventing different kinds of regularizers. Extensive experiments on both synthetic and real data sets show the effectiveness of our method.

Original languageEnglish
Pages (from-to)567-597
Number of pages31
JournalKnowledge and Information Systems
Volume38
Issue number3
DOIs
Publication statusPublished - Jan 1 2014

Fingerprint

Labels
Experiments

All Science Journal Classification (ASJC) codes

  • Software
  • Information Systems
  • Human-Computer Interaction
  • Hardware and Architecture
  • Artificial Intelligence

Cite this

Transfer dimensionality reduction by Gaussian process in parallel. / Tong, Bin; Gao, Junbin; Nguyen Huy, Thach; Shao, Hao; Suzuki, Einoshin.

In: Knowledge and Information Systems, Vol. 38, No. 3, 01.01.2014, p. 567-597.

Research output: Contribution to journalArticle

Tong, Bin ; Gao, Junbin ; Nguyen Huy, Thach ; Shao, Hao ; Suzuki, Einoshin. / Transfer dimensionality reduction by Gaussian process in parallel. In: Knowledge and Information Systems. 2014 ; Vol. 38, No. 3. pp. 567-597.
@article{749740f2525e4b5db9d2dfa6890bf3f9,
title = "Transfer dimensionality reduction by Gaussian process in parallel",
abstract = "Dimensionality reduction has been considered as one of the most significant tools for data analysis. In general, supervised information is helpful for dimensionality reduction. However, in typical real applications, supervised information in multiple source tasks may be available, while the data of the target task are unlabeled. An interesting problem of how to guide the dimensionality reduction for the unlabeled target data by exploiting useful knowledge, such as label information, from multiple source tasks arises in such a scenario. In this paper, we propose a new method for dimensionality reduction in the transfer learning setting. Unlike traditional paradigms where the useful knowledge from multiple source tasks is transferred through distance metric, we attempt to learn a more informative mapping function between the original data and the reduced data by Gaussian process that behaves more appropriately than other parametric regression methods due to its less parametric characteristic. In our proposal, we firstly convert the dimensionality reduction problem into integral regression problems in parallel. Gaussian process is then employed to learn the underlying relationship between the original data and the reduced data. Such a relationship can be appropriately transferred to the target task by exploiting the prediction ability of the Gaussian process model and inventing different kinds of regularizers. Extensive experiments on both synthetic and real data sets show the effectiveness of our method.",
author = "Bin Tong and Junbin Gao and {Nguyen Huy}, Thach and Hao Shao and Einoshin Suzuki",
year = "2014",
month = "1",
day = "1",
doi = "10.1007/s10115-012-0601-y",
language = "English",
volume = "38",
pages = "567--597",
journal = "Knowledge and Information Systems",
issn = "0219-1377",
publisher = "Springer London",
number = "3",

}

TY - JOUR

T1 - Transfer dimensionality reduction by Gaussian process in parallel

AU - Tong, Bin

AU - Gao, Junbin

AU - Nguyen Huy, Thach

AU - Shao, Hao

AU - Suzuki, Einoshin

PY - 2014/1/1

Y1 - 2014/1/1

N2 - Dimensionality reduction has been considered as one of the most significant tools for data analysis. In general, supervised information is helpful for dimensionality reduction. However, in typical real applications, supervised information in multiple source tasks may be available, while the data of the target task are unlabeled. An interesting problem of how to guide the dimensionality reduction for the unlabeled target data by exploiting useful knowledge, such as label information, from multiple source tasks arises in such a scenario. In this paper, we propose a new method for dimensionality reduction in the transfer learning setting. Unlike traditional paradigms where the useful knowledge from multiple source tasks is transferred through distance metric, we attempt to learn a more informative mapping function between the original data and the reduced data by Gaussian process that behaves more appropriately than other parametric regression methods due to its less parametric characteristic. In our proposal, we firstly convert the dimensionality reduction problem into integral regression problems in parallel. Gaussian process is then employed to learn the underlying relationship between the original data and the reduced data. Such a relationship can be appropriately transferred to the target task by exploiting the prediction ability of the Gaussian process model and inventing different kinds of regularizers. Extensive experiments on both synthetic and real data sets show the effectiveness of our method.

AB - Dimensionality reduction has been considered as one of the most significant tools for data analysis. In general, supervised information is helpful for dimensionality reduction. However, in typical real applications, supervised information in multiple source tasks may be available, while the data of the target task are unlabeled. An interesting problem of how to guide the dimensionality reduction for the unlabeled target data by exploiting useful knowledge, such as label information, from multiple source tasks arises in such a scenario. In this paper, we propose a new method for dimensionality reduction in the transfer learning setting. Unlike traditional paradigms where the useful knowledge from multiple source tasks is transferred through distance metric, we attempt to learn a more informative mapping function between the original data and the reduced data by Gaussian process that behaves more appropriately than other parametric regression methods due to its less parametric characteristic. In our proposal, we firstly convert the dimensionality reduction problem into integral regression problems in parallel. Gaussian process is then employed to learn the underlying relationship between the original data and the reduced data. Such a relationship can be appropriately transferred to the target task by exploiting the prediction ability of the Gaussian process model and inventing different kinds of regularizers. Extensive experiments on both synthetic and real data sets show the effectiveness of our method.

UR - http://www.scopus.com/inward/record.url?scp=84894646808&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84894646808&partnerID=8YFLogxK

U2 - 10.1007/s10115-012-0601-y

DO - 10.1007/s10115-012-0601-y

M3 - Article

AN - SCOPUS:84894646808

VL - 38

SP - 567

EP - 597

JO - Knowledge and Information Systems

JF - Knowledge and Information Systems

SN - 0219-1377

IS - 3

ER -