Semi-Supervised Domain Adaptation via Minimax Entropy

Saito, Kuniaki; Kim, Donghyun; Sclaroff, Stan; Darrell, Trevor; Saenko, Kate

doi:10.1109/iccv.2019.00814

Cited by 569 publications

(621 citation statements)

References 18 publications

Supporting

Mentioning

570

Contrasting

Order By: Relevance

“…However, as mentioned, even if global domain distribution alignment is enforced, it often leads to per-class alignment, which reduces the discriminativeness of the learned feature representation for the FSL task. Moreover, since existing UDA methods still assume that the target domain contains the same classes as the source domain, the more recent methods that focus on per-class cross-domain alignment [25]- [29] are unsuitable for our CD-FSL problem. Thus, global domain data distribution alignment [9], [14], [47] is adopted in our DPDAPN with a special mechanism introduced to prevent per-class alignment.…”

Section: Related Workmentioning

confidence: 99%

“…Nevertheless, a naïve combination of existing DA and FSL methods fails to offer an effective solution (see Tables 1&2) because the existing UDA methods assume that the target and source domains have identical label space. Given that they are mainly designed for distribution alignment across domains (recently focusing on per-class alignment [25]- [29]), they are intrinsically unsuited for FSL whereby the target classes are completely different from the source classes; either global or per-class distribution alignment would have a detrimental effect on class separation and model discriminativeness. How to achieve domain distribution alignment for DA while maintaining source/target perclass discriminativeness thus becomes the key to CD-FSL.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Discriminativeness-Preserved Domain Adaptation for Few-Shot Learning

Liu

2020

IEEE Access

View full text Add to dashboard Cite

Existing few-shot learning (FSL) methods make the implicit assumption that the few target class samples are from the same domain as the source class samples. However, this assumption is often invalid in practice-the target classes could come from a different domain. This poses an additional domain adaptation (DA) challenge with few training samples. In this paper, the problem of cross-domain few-shot learning (CD-FSL) is approached, which requires solving FSL and DA in a unified framework. To this end, we propose a novel discriminativeness-preserved domain adaptive prototypical network (DPDAPN) model. It is designed to address a specific challenge in CD-FSL: the DA objective means that the source and target data distributions need to be aligned, typically through a shared domain adaptive feature embedding space, but the FSL objective dictates that the target domain per-class distribution must be different from that of any source domain class, meaning aligning the distributions across domains may harm the FSL performance. How to achieve global domain distribution alignment while maintaining source/target perclass discriminativeness thus becomes the key. Our solution is to explicitly enhance the source/target perclass separation before domain adaptive feature embedding learning in DPDAPN to alleviate the negative effect of domain alignment on FSL. Extensive experiments show that our DPDAPN outperforms the stateof-the-art FSL and DA models, as well as their naive combinations. INDEX TERMS Discriminativeness, domain adaptation, few-shot learning, prototypical network.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Discriminativeness-Preserved Domain Adaptation for Few-Shot Learning

Liu

2020

IEEE Access

View full text Add to dashboard Cite

show abstract

“…A plethora of SSDA works [18,25,36] have already emerged. For example, the [36] designs a min-max entropy minimization strategy to achieve better adaptation.…”

Section: Related Workmentioning

confidence: 99%

“…A plethora of SSDA works [18,25,36] have already emerged. For example, the [36] designs a min-max entropy minimization strategy to achieve better adaptation. Some works [18,25] consider the intra-adaptation bias but they either leverage the labeled data to learn discriminative features like [18] or minimize the entropy similarity between intra-target samples like [25].…”

Section: Related Workmentioning

confidence: 99%

Alleviating Human-level Shift

Xu¹,

Zou²,

Lin³

2020

Proceedings of the 28th ACM International Conference on Multimedia

View full text Add to dashboard Cite

Human pose estimation has been widely studied with much focus on supervised learning requiring sufficient annotations. However, in real applications, a pretrained pose estimation model usually need be adapted to a novel domain with no labels or sparse labels. Such domain adaptation for 2D pose estimation hasn't been explored. The main reason is that a pose, by nature, has typical topological structure and needs fine-grained features in local keypoints. While existing adaptation methods do not consider topological structure of object-of-interest and they align the whole images coarsely. Therefore, we propose a novel domain adaptation method for multi-person pose estimation to conduct the human-level topological structure alignment and fine-grained feature alignment. Our method consists of three modules: Cross-Attentive Feature Alignment (CAFA), Intra-domain Structure Adaptation (ISA) and Inter-domain Human-Topology Alignment (IHTA) module. The CAFA adopts a bidirectional spatial attention module (BSAM) that focuses on fine-grained local feature correlation between two humans to adaptively aggregate consistent features for adaptation. We adopt ISA only in semi-supervised domain adaptation (SSDA) to exploit the corresponding keypoint semantic relationship for reducing the intra-domain bias. Most importantly, we propose an IHTA to learn more domain-invariant human topological representation for reducing the inter-domain discrepancy. We model the human topological structure via the graph convolution network (GCN), by passing messages on which, high-order relations can be considered. This structure preserving alignment based on GCN is beneficial to the occluded or extreme pose inference. Extensive experiments are conducted on two popular benchmarks and results demonstrate the competency of our method compared with existing supervised approaches. CCS CONCEPTS • Computing methodologies → Activity recognition and understanding; Transfer learning;

show abstract

“…For the semi-supervised methods, labeled source domain, a small number of annotated target samples, along with unlabeled target samples, are used in the training process. The proposed method in this paper subsumes under the umbrella of the semi-supervised HDA, in light of the prior works, such as [2,21,27], have demonstrated the effectiveness of exploiting the unlabeled target domain data.…”

Section: Introductionmentioning

confidence: 98%

Prototype-Matching Graph Network for Heterogeneous Domain Adaptation

Wang

Luo

Huang

et al. 2020

Proceedings of the 28th ACM International Conference on Multimedia

View full text Add to dashboard Cite

Even though the multimedia data is ubiquitous on the web, the scarcity of the annotated data and variety of data modalities hinder their usage by multimedia applications. Heterogeneous domain adaptation (HDA) has therefore arisen to address such limitations by facilitating the knowledge transfer between heterogeneous domains. Existing HDA methods only focus on aligning the crossdomain feature distributions and ignore the importance of maximizing the margin among different classes, which may lead to a suboptimal classification performance. To tackle this problem, in this paper, we propose the Prototype-Matching Graph Network (PMGN), which gradually explores the domain-invariant class prototype representations. Specifically, we build an end-to-end Graph Prototypical Network, which computes the class prototypes through multiple layers of edge learning, node aggregation, and discrepancy minimization. Our framework utilizes the Swap training strategy to provide adequate supervision for training the edge learning component. Moreover, the proposed PMGN can be equipped with the clustering module that utilises the KL-divergence as a distance metric to reduce the distribution difference between the source and target data. Extensive experiments on three HDA tasks (i.e. object recognition, text-to-image classification, and text categorization) demonstrate the superiority of our approach over the state-of-theart HDA methods. CCS CONCEPTS • Computing methodologies → Transfer learning.

show abstract

Semi-Supervised Domain Adaptation via Minimax Entropy

Cited by 569 publications

References 18 publications

Discriminativeness-Preserved Domain Adaptation for Few-Shot Learning

Discriminativeness-Preserved Domain Adaptation for Few-Shot Learning

Alleviating Human-level Shift

Prototype-Matching Graph Network for Heterogeneous Domain Adaptation

Contact Info

Product

Resources

About