Cycle Consistency Based Pseudo Label and Fine Alignment for Unsupervised Domain Adaptation

Zhang, Hui; Tang, Junkun; Cao, Yihong; Chen, Yurong; Wang, Yaonan; Wu, Q. M. Jonathan

doi:10.1109/tmm.2022.3233306

Cited by 6 publications

(8 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Some metric-based UDA schemes utilize special mechanisms of adapting source and target features. The cycle-consistent loss is developed in [7] and [8]. The target features is also capable of forecasting the class the source features belong to if the source features can be utilized to predict the class the target features belong to.…”

Section: A Unsupervised Domain Adaptionmentioning

confidence: 99%

“…where 𝑝 is the prediction of the 𝑖th classifier. Using (3), ( 5) and ( 6), Property I can be derived and contrastive discrepancy loss ℓ can be formulated in (7). More details of the derivations can be found in Appendix A.…”

Section: 𝐻(𝑝 𝑝 ) = 𝔼 ~ [− Log 𝑝 (𝑦|𝒙 )𝑝 (𝑦|𝒙 )]mentioning

confidence: 99%

“…where 𝑀 = 𝑁 + 𝑁 , 𝜏 and 𝒙 ( ) express the number of negative samples, the temperature coefficient, and a set of negative samples, respectively. To reduce computational complexity for (7), this paper adopts the nearest 𝑐 for 𝑓…”

Section: 𝐻(𝑝 𝑝 ) = 𝔼 ~ [− Log 𝑝 (𝑦|𝒙 )𝑝 (𝑦|𝒙 )]mentioning

confidence: 99%

“…rather than computing the average of 𝒄 . Comparing (7) with (3), ℓ in (7) leverages more information between anchor features, 𝑓 and 𝑓 , and negative features, 𝒇 ( ) for 𝒙 ( ) . Additionally, the indicator function, 1 ( ) , , is zero when these two classifiers have predictions, 𝑦 ( ) = 𝑘 and 𝑦 = 𝑘 .…”

Section: > Replace This Line With Your Manuscript Id Number (Double-c...mentioning

confidence: 99%

“…Metric learning-based methods strengthen the relationship between positive and negative features to make the feature clusters more distinguishable. Cycle-consistent-based schemes assume the centers of target features of classes can be used to forecast the categories of source features [7,8]. Another branch of the kind of class-conditioned UDA is called the bi-R 2 > REPLACE THIS LINE WITH YOUR MANUSCRIPT ID NUMBER (DOUBLE-CLICK HERE TO EDIT) <  classifier method.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

The Enhancement of Classification of Imbalanced Dataset for Edge Computing

Huang

Hsu

Hung

et al. 2022

Communications in Computer and Information Science

View full text Add to dashboard Cite

This paper proposes a novel unsupervised domain adaption (UDA) method based on contrastive bi-projector (CBP), which can improve the existing UDA methods. It is called CBPUDA here, which effectively promotes the feature extractors (FEs) to reduce the generation of ambiguous features for classification and domain adaption. The CBP differs from traditional bi-classifier-based methods at that these two classifiers are replaced with two projectors of performing a mapping from the input feature to two distinct features. These two projectors and the FEs in the CBPUDA can be trained adversarially to obtain more refined decision boundaries so that it can possess powerful classification performance. Two properties of the proposed loss function are analyzed here. The first property is to derive an upper bound of joint prediction entropy, which is used to form the proposed loss function, contrastive discrepancy (CD) loss. The CD loss takes the advantages of the contrastive learning and the bi-classifier. The second property is to analyze the gradient of the CD loss and then overcome the drawback of the CD loss. The result of the second property is utilized in the development of the gradient scaling (GS) scheme in this paper. The GS scheme can be exploited to tackle the unstable problem of the CD loss because training the CBPUDA requires using contrastive learning and adversarial learning at the same time. Therefore, using the CD loss with the GS scheme overcomes the problem mentioned above to make features more compact for intra-class and distinguishable for inter-class. Experimental results express that the CBPUDA is superior to conventional UDA methods under consideration in this paper for UDA and fine-grained UDA tasks.

show abstract

Section: A Unsupervised Domain Adaptionmentioning

confidence: 99%

Section: 𝐻(𝑝 𝑝 ) = 𝔼 ~ [− Log 𝑝 (𝑦|𝒙 )𝑝 (𝑦|𝒙 )]mentioning

confidence: 99%

Section: 𝐻(𝑝 𝑝 ) = 𝔼 ~ [− Log 𝑝 (𝑦|𝒙 )𝑝 (𝑦|𝒙 )]mentioning

confidence: 99%

Section: > Replace This Line With Your Manuscript Id Number (Double-c...mentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations