Dual transfer learning

Mingsheng Long, Jianmin Wang, Guiguang Ding, Wei Cheng, Xiang Zhang, Wei Wang

Research output: Chapter in Book/Report/Conference proceedingConference contribution

36 Citations (Scopus)

Abstract

Transfer learning aims to leverage the knowledge in the source domain to facilitate the learning tasks in the tar- get domain. It has attracted extensive research interests recently due to its effectiveness in a wide range of applica- Tions. The general idea of the existing methods is to utilize the common latent structure shared across domains as the bridge for knowledge transfer. These methods usually model the common latent structure by using either the marginal distribution or the conditional distribution. However, with- out exploring the duality between these two distributions, these single bridge methods may not achieve optimal capa- bility of knowledge transfer. In this paper, we propose a novel approach, Dual Transfer Learning (DTL), which simultaneously learns the marginal and conditional distributions, and exploits the du- Ality between them in a principled way. The key idea behind DTL is that learning one distribution can help to learn the other. This duality property leads to mutual reinforcement when adapting both distributions across domains to trans- fer knowledge. The proposed method is formulated as an optimization problem based on joint nonnegative matrix tri- factorizations (NMTF). The two distributions are learned from the decomposed latent factors that exhibit the duality property. An efficient alternating minimization algorithm is developed to solve the optimization problem with con- vergence guarantee. Extensive experimental results demon- strate that DTL is more effective than alternative transfer learning methods.

Original languageEnglish (US)
Title of host publicationProceedings of the 12th SIAM International Conference on Data Mining, SDM 2012
Pages540-551
Number of pages12
StatePublished - Dec 1 2012
Event12th SIAM International Conference on Data Mining, SDM 2012 - Anaheim, CA, United States
Duration: Apr 26 2012Apr 28 2012

Other

Other12th SIAM International Conference on Data Mining, SDM 2012
CountryUnited States
CityAnaheim, CA
Period4/26/124/28/12

Fingerprint

Factorization
Reinforcement

All Science Journal Classification (ASJC) codes

  • Computer Science Applications

Cite this

Long, M., Wang, J., Ding, G., Cheng, W., Zhang, X., & Wang, W. (2012). Dual transfer learning. In Proceedings of the 12th SIAM International Conference on Data Mining, SDM 2012 (pp. 540-551)
Long, Mingsheng ; Wang, Jianmin ; Ding, Guiguang ; Cheng, Wei ; Zhang, Xiang ; Wang, Wei. / Dual transfer learning. Proceedings of the 12th SIAM International Conference on Data Mining, SDM 2012. 2012. pp. 540-551
@inproceedings{ef464ef4f686477ebd14f5136db6c4d1,
title = "Dual transfer learning",
abstract = "Transfer learning aims to leverage the knowledge in the source domain to facilitate the learning tasks in the tar- get domain. It has attracted extensive research interests recently due to its effectiveness in a wide range of applica- Tions. The general idea of the existing methods is to utilize the common latent structure shared across domains as the bridge for knowledge transfer. These methods usually model the common latent structure by using either the marginal distribution or the conditional distribution. However, with- out exploring the duality between these two distributions, these single bridge methods may not achieve optimal capa- bility of knowledge transfer. In this paper, we propose a novel approach, Dual Transfer Learning (DTL), which simultaneously learns the marginal and conditional distributions, and exploits the du- Ality between them in a principled way. The key idea behind DTL is that learning one distribution can help to learn the other. This duality property leads to mutual reinforcement when adapting both distributions across domains to trans- fer knowledge. The proposed method is formulated as an optimization problem based on joint nonnegative matrix tri- factorizations (NMTF). The two distributions are learned from the decomposed latent factors that exhibit the duality property. An efficient alternating minimization algorithm is developed to solve the optimization problem with con- vergence guarantee. Extensive experimental results demon- strate that DTL is more effective than alternative transfer learning methods.",
author = "Mingsheng Long and Jianmin Wang and Guiguang Ding and Wei Cheng and Xiang Zhang and Wei Wang",
year = "2012",
month = "12",
day = "1",
language = "English (US)",
isbn = "9781611972320",
pages = "540--551",
booktitle = "Proceedings of the 12th SIAM International Conference on Data Mining, SDM 2012",

}

Long, M, Wang, J, Ding, G, Cheng, W, Zhang, X & Wang, W 2012, Dual transfer learning. in Proceedings of the 12th SIAM International Conference on Data Mining, SDM 2012. pp. 540-551, 12th SIAM International Conference on Data Mining, SDM 2012, Anaheim, CA, United States, 4/26/12.

Dual transfer learning. / Long, Mingsheng; Wang, Jianmin; Ding, Guiguang; Cheng, Wei; Zhang, Xiang; Wang, Wei.

Proceedings of the 12th SIAM International Conference on Data Mining, SDM 2012. 2012. p. 540-551.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Dual transfer learning

AU - Long, Mingsheng

AU - Wang, Jianmin

AU - Ding, Guiguang

AU - Cheng, Wei

AU - Zhang, Xiang

AU - Wang, Wei

PY - 2012/12/1

Y1 - 2012/12/1

N2 - Transfer learning aims to leverage the knowledge in the source domain to facilitate the learning tasks in the tar- get domain. It has attracted extensive research interests recently due to its effectiveness in a wide range of applica- Tions. The general idea of the existing methods is to utilize the common latent structure shared across domains as the bridge for knowledge transfer. These methods usually model the common latent structure by using either the marginal distribution or the conditional distribution. However, with- out exploring the duality between these two distributions, these single bridge methods may not achieve optimal capa- bility of knowledge transfer. In this paper, we propose a novel approach, Dual Transfer Learning (DTL), which simultaneously learns the marginal and conditional distributions, and exploits the du- Ality between them in a principled way. The key idea behind DTL is that learning one distribution can help to learn the other. This duality property leads to mutual reinforcement when adapting both distributions across domains to trans- fer knowledge. The proposed method is formulated as an optimization problem based on joint nonnegative matrix tri- factorizations (NMTF). The two distributions are learned from the decomposed latent factors that exhibit the duality property. An efficient alternating minimization algorithm is developed to solve the optimization problem with con- vergence guarantee. Extensive experimental results demon- strate that DTL is more effective than alternative transfer learning methods.

AB - Transfer learning aims to leverage the knowledge in the source domain to facilitate the learning tasks in the tar- get domain. It has attracted extensive research interests recently due to its effectiveness in a wide range of applica- Tions. The general idea of the existing methods is to utilize the common latent structure shared across domains as the bridge for knowledge transfer. These methods usually model the common latent structure by using either the marginal distribution or the conditional distribution. However, with- out exploring the duality between these two distributions, these single bridge methods may not achieve optimal capa- bility of knowledge transfer. In this paper, we propose a novel approach, Dual Transfer Learning (DTL), which simultaneously learns the marginal and conditional distributions, and exploits the du- Ality between them in a principled way. The key idea behind DTL is that learning one distribution can help to learn the other. This duality property leads to mutual reinforcement when adapting both distributions across domains to trans- fer knowledge. The proposed method is formulated as an optimization problem based on joint nonnegative matrix tri- factorizations (NMTF). The two distributions are learned from the decomposed latent factors that exhibit the duality property. An efficient alternating minimization algorithm is developed to solve the optimization problem with con- vergence guarantee. Extensive experimental results demon- strate that DTL is more effective than alternative transfer learning methods.

UR - http://www.scopus.com/inward/record.url?scp=84868294775&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84868294775&partnerID=8YFLogxK

M3 - Conference contribution

SN - 9781611972320

SP - 540

EP - 551

BT - Proceedings of the 12th SIAM International Conference on Data Mining, SDM 2012

ER -

Long M, Wang J, Ding G, Cheng W, Zhang X, Wang W. Dual transfer learning. In Proceedings of the 12th SIAM International Conference on Data Mining, SDM 2012. 2012. p. 540-551