TY - JOUR
T1 - Heterogeneous Domain Adaptation by Information Capturing and Distribution Matching
AU - Wu, Hanrui
AU - Zhu, Hong
AU - Yan, Yuguang
AU - Wu, Jiaju
AU - Zhang, Yifan
AU - Ng, Michael K.
N1 - This work was supported in part by the Research Grants Council (RGC) of Hong Kong, General Research Fund (HKRGC GRF) under Grant 12200317, Grant 12300218, Grant 12300519, and Grant 17201020; and in part by the HKU–TCL Joint Research Centre for Artificial Intelligence, Hong Kong. The work of Hong Zhu was supported in part by the NSF of China under Grant NSFC 11701227 and Grant NSFC 11971149, in part by the NSF of Jiangsu Province under Project BK20170522, and in part by Jiangsu University under Grant 17JDG013.
Publisher Copyright:
© 1992-2012 IEEE.
PY - 2021/7/8
Y1 - 2021/7/8
N2 - Heterogeneous domain adaptation (HDA) is a challenging problem because of the different feature representations in the source and target domains. Most HDA methods search for mapping matrices from the source and target domains to discover latent features for learning. However, these methods barely consider the reconstruction error to measure the information loss during the mapping procedure. In this paper, we propose to jointly capture the information and match the source and target domain distributions in the latent feature space. In the learning model, we propose to minimize the reconstruction loss between the original and reconstructed representations to preserve information during transformation and reduce the Maximum Mean Discrepancy between the source and target domains to align their distributions. The resulting minimization problem involves two projection variables with orthogonal constraints that can be solved by the generalized gradient flow method, which can preserve orthogonal constraints in the computational procedure. We conduct extensive experiments on several image classification datasets to demonstrate that the effectiveness and efficiency of the proposed method are better than those of state-of-the-art HDA methods.
AB - Heterogeneous domain adaptation (HDA) is a challenging problem because of the different feature representations in the source and target domains. Most HDA methods search for mapping matrices from the source and target domains to discover latent features for learning. However, these methods barely consider the reconstruction error to measure the information loss during the mapping procedure. In this paper, we propose to jointly capture the information and match the source and target domain distributions in the latent feature space. In the learning model, we propose to minimize the reconstruction loss between the original and reconstructed representations to preserve information during transformation and reduce the Maximum Mean Discrepancy between the source and target domains to align their distributions. The resulting minimization problem involves two projection variables with orthogonal constraints that can be solved by the generalized gradient flow method, which can preserve orthogonal constraints in the computational procedure. We conduct extensive experiments on several image classification datasets to demonstrate that the effectiveness and efficiency of the proposed method are better than those of state-of-the-art HDA methods.
KW - domain adaptation
KW - generalized gradient flow
KW - Heterogeneous domain adaptation
KW - transfer learning
UR - http://www.scopus.com/inward/record.url?scp=85110457649&partnerID=8YFLogxK
U2 - 10.1109/TIP.2021.3094137
DO - 10.1109/TIP.2021.3094137
M3 - Journal article
C2 - 34236965
AN - SCOPUS:85110457649
SN - 1057-7149
VL - 30
SP - 6364
EP - 6376
JO - IEEE Transactions on Image Processing
JF - IEEE Transactions on Image Processing
ER -