東華大學圖書館 |

Language: English

Help

回圖書館首頁

手機版館藏查詢

Back

Switch To: Labeled | MARC Mode | ISBD

Domain Adaptation through Optimal Tr...

Gao, Keyue.

Linked to FindBook

Google Book

Amazon

博客來

Domain Adaptation through Optimal Transport with Class Imbalance.

Record Type:	Electronic resources : Monograph/item
Title/Author:	Domain Adaptation through Optimal Transport with Class Imbalance./
Author:	Gao, Keyue.
Published:	Ann Arbor : ProQuest Dissertations & Theses, : 2018,
Description:	71 p.
Notes:	Source: Dissertations Abstracts International, Volume: 80-05, Section: B.
Contained By:	Dissertations Abstracts International80-05B.
Subject:	Applied Mathematics. -
Online resource:	http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10751110
ISBN:	9780438634558

Domain Adaptation through Optimal Transport with Class Imbalance.
Gao, Keyue.

Domain Adaptation through Optimal Transport with Class Imbalance. - Ann Arbor : ProQuest Dissertations & Theses, 2018 - 71 p.

Source: Dissertations Abstracts International, Volume: 80-05, Section: B.

Thesis (Ph.D.)--New York University, 2018.

This item must not be sold to any third party vendors.

In modern data analytics, one often faces situations where the training data used to build a model and the target data to which the model is applied have different but related probability distributions. The problem of learning effectively in such scenarios is known as domain adaptation. In this paper, we propose a new framework to solve the domain adaptation problem based on optimal transport. Our approach removes two assumptions underlying prior related work. First, we allow class imbalance: that the class ratios in the training and the target data be different. Second, we allow that only a subdomain of the training data be matched to the target data, a relaxation that we name optimal partial transport. We introduce various techniques in optimal transport to address these two situations. We also present a novel regularization method which promotes small variance within each class. Numerical experiments are conducted on both synthetic and real world data, showing that our approach is effective and robust with fewer assumptions made on data than with conventional methods.

ISBN: 9780438634558Subjects--Topical Terms:

1669109
Applied Mathematics.

Domain Adaptation through Optimal Transport with Class Imbalance.
LDR:02137nmm a2200325 4500 001 2209605
005 20191104073749.5
008 201008s2018 ||||||||||||||||| ||eng d
020 $a 9780438634558
035 $a (MiAaPQ)AAI10751110
035 $a (MiAaPQ)nyu:13303
035 $a AAI10751110
040 $a MiAaPQ $c MiAaPQ
100 1 $a Gao, Keyue. $3 3436702
245 1 0 $a Domain Adaptation through Optimal Transport with Class Imbalance.
260 1 $a Ann Arbor : $b ProQuest Dissertations & Theses, $c 2018
300 $a 71 p.
500 $a Source: Dissertations Abstracts International, Volume: 80-05, Section: B.
500 $a Publisher info.: Dissertation/Thesis.
500 $a Advisor: Tabak, Esteban.
502 $a Thesis (Ph.D.)--New York University, 2018.
506 $a This item must not be sold to any third party vendors.
520 $a In modern data analytics, one often faces situations where the training data used to build a model and the target data to which the model is applied have different but related probability distributions. The problem of learning effectively in such scenarios is known as domain adaptation. In this paper, we propose a new framework to solve the domain adaptation problem based on optimal transport. Our approach removes two assumptions underlying prior related work. First, we allow class imbalance: that the class ratios in the training and the target data be different. Second, we allow that only a subdomain of the training data be matched to the target data, a relaxation that we name optimal partial transport. We introduce various techniques in optimal transport to address these two situations. We also present a novel regularization method which promotes small variance within each class. Numerical experiments are conducted on both synthetic and real world data, showing that our approach is effective and robust with fewer assumptions made on data than with conventional methods.
590 $a School code: 0146.
650 4 $a Applied Mathematics. $3 1669109
650 4 $a Mathematics. $3 515831
690 $a 0364
690 $a 0405
710 2 $a New York University. $b Mathematics. $3 1019424
773 0 $t Dissertations Abstracts International $g 80-05B.
790 $a 0146
791 $a Ph.D.
792 $a 2018
793 $a English
856 4 0 $u http://pqdd.sinica.edu.tw/twdaoapp/servlet/advanced?query=10751110