TY - JOUR
T1 - Unsupervised domain adaptation with adversarial residual transform networks
AU - Cai, Guanyu
AU - Wang, Yuqin
AU - He, Lianghua
AU - Zhou, Mengchu
N1 - Funding Information:
Manuscript received July 16, 2018; revised December 11, 2018 and May 19, 2019; accepted August 2, 2019. Date of publication September 11, 2019; date of current version August 4, 2020. This work was supported in part by the National Natural Science Foundation of China under Grant 61772369, Grant 61773166, Grant 61771144, and Grant 61871004, in part by Joint Funds of the National Science Foundation of China under Grant U18092006, in part by the Shanghai Municipal Science and Technology Committee of Shanghai Outstanding Academic Leaders Plan under Grant 19XD1434000, in part by the Projects of International Cooperation of Shanghai Municipal Science and Technology Committee under Grant 19490712800, in part by the Shanghai Science and Technology Committee under Grant 17411953100, in part by the National Key Research and Development Program under Grant 2018YFB1004701, and in part by the Fundamental Research Funds for the Central Universities. (Corresponding authors: Lianghua He; MengChu Zhou.) G. Cai, Y. Wang, and L. He are with the Department of Computer Science and Technology, Tongji University, Shanghai 201804, China (e-mail: caiguanyu@tongji.edu.cn; wangyuqin@tongji.edu.cn; helianghua@ tongji.edu.cn).
Publisher Copyright:
© 2012 IEEE.
PY - 2020/8
Y1 - 2020/8
N2 - Domain adaptation (DA) is widely used in learning problems lacking labels. Recent studies show that deep adversarial DA models can make markable improvements in performance, which include symmetric and asymmetric architectures. However, the former has poor generalization ability, whereas the latter is very hard to train. In this article, we propose a novel adversarial DA method named adversarial residual transform networks (ARTNs) to improve the generalization ability, which directly transforms the source features into the space of target features. In this model, residual connections are used to share features and adversarial loss is reconstructed, thus making the model more generalized and easier to train. Moreover, a special regularization term is added to the loss function to alleviate a vanishing gradient problem, which enables its training process stable. A series of experiments based on Amazon review data set, digits data sets, and Office-31 image data sets are conducted to show that the proposed ARTN can be comparable with the methods of the state of the art.
AB - Domain adaptation (DA) is widely used in learning problems lacking labels. Recent studies show that deep adversarial DA models can make markable improvements in performance, which include symmetric and asymmetric architectures. However, the former has poor generalization ability, whereas the latter is very hard to train. In this article, we propose a novel adversarial DA method named adversarial residual transform networks (ARTNs) to improve the generalization ability, which directly transforms the source features into the space of target features. In this model, residual connections are used to share features and adversarial loss is reconstructed, thus making the model more generalized and easier to train. Moreover, a special regularization term is added to the loss function to alleviate a vanishing gradient problem, which enables its training process stable. A series of experiments based on Amazon review data set, digits data sets, and Office-31 image data sets are conducted to show that the proposed ARTN can be comparable with the methods of the state of the art.
KW - Adversarial neural networks
KW - residual connections
KW - transfer learning
KW - unsupervised domain adaptation (DA)
UR - http://www.scopus.com/inward/record.url?scp=85089124828&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85089124828&partnerID=8YFLogxK
U2 - 10.1109/TNNLS.2019.2935384
DO - 10.1109/TNNLS.2019.2935384
M3 - Article
C2 - 31514161
AN - SCOPUS:85089124828
SN - 2162-237X
VL - 31
SP - 3073
EP - 3086
JO - IEEE Transactions on Neural Networks and Learning Systems
JF - IEEE Transactions on Neural Networks and Learning Systems
IS - 8
M1 - 8833506
ER -