Paper Info
Title
Informative pairs mining based adaptive metric learning for adversarial domain adaptation
Abstract
Adversarial domain adaptation has made remarkable in promoting feature transferability, while recent work reveals that there exists an unexpected degradation of feature discrimination during the procedure of learning transferable features. This paper proposes an informative pairs mining based adaptive metric learning (IPM-AML), where a novel two-triplet-sampling strategy is advanced to select informative positive pairs from the same classes and informative negative pairs from different classes, and a metric loss imposed with special weights is further utilized to adaptively pay more attention to those more informative pairs which can adaptively improve discrimination. Then, we incorporate IPM-AML into popular conditional domain adversarial network (CDAN) to learn feature representation that is transferable and discriminative desirably (IPM-AML-CDAN). To ensure the reliability of pseudo target labels in the whole training process, we select more confident target ones whose predicted scores are higher than a given threshold T, and also provide theoretical validation for this simple threshold strategy. Extensive experiment results on four cross-domain benchmarks validate that IPM-AML-CDAN can achieve competitive results compared with state-of-the-art approaches.
Year
DOI
Venue
2022
10.1016/j.neunet.2022.03.031
Neural Networks
Keywords
DocType
Volume
Domain adaptation,Informative pairs mining,Adaptive metric learning,Adversarial domain adaptation
Journal
151
ISSN
Citations 
PageRank 
0893-6080
0
0.34
References 
Authors
0
9
Name
Order
Citations
PageRank
Mengzhu Wang123.44
Paul Li200.68
Li Shen3863102.99
H. Y. Wang410018.32
Shanshan Wang515027.21
Wei Wang600.34
Xiang Zhang700.34
Junyang Chen8386.68
Zhigang Luo982847.92