Deep Residual Correction Network for Partial Domain Adaptation

Li, Shuang; Liu, Chi Harold; Lin, Qiuxia; Qi, Wen; Su, Limin; Huang, Gao; Ding, Zhengming

doi:10.1109/tpami.2020.2964173

Cited by 132 publications

(79 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Another example is RTN [38] that considers feature fusion with MMD and designs a residual function to perform classifier adaptation. Further, DRCN in [24] utilizes residual correction block to explicitly mitigate the domain feature gap. Apart from MMD, Zhang et al [22] define a new divergence, Margin Disparity Discrepancy (MDD), and validate that it has rigorous generalization bounds.…”

Section: Discrepancy Metric Minimizationmentioning

confidence: 99%

See 1 more Smart Citation

Domain Conditioned Adaptation Network

Liu

Lin

et al. 2020

AAAI

Self Cite

View full text Add to dashboard Cite

Tremendous research efforts have been made to thrive deep domain adaptation (DA) by seeking domain-invariant features. Most existing deep DA models only focus on aligning feature representations of task-specific layers across domains while integrating a totally shared convolutional architecture for source and target. However, we argue that such strongly-shared convolutional layers might be harmful for domain-specific feature learning when source and target data distribution differs to a large extent. In this paper, we relax a shared-convnets assumption made by previous DA methods and propose a Domain Conditioned Adaptation Network (DCAN), which aims to excite distinct convolutional channels with a domain conditioned channel attention mechanism. As a result, the critical low-level domain-dependent knowledge could be explored appropriately. As far as we know, this is the first work to explore the domain-wise convolutional channel activation for deep DA networks. Moreover, to effectively align high-level feature distributions across two domains, we further deploy domain conditioned feature correction blocks after task-specific layers, which will explicitly correct the domain discrepancy. Extensive experiments on three cross-domain benchmarks demonstrate the proposed approach outperforms existing methods by a large margin, especially on very tough cross-domain learning tasks.

show abstract

Section: Discrepancy Metric Minimizationmentioning

confidence: 99%

“…Unlike previous works [24], [38], we adapt all taskspecific layers, and more importantly, explicitly measure domain discrepancy from a structural aspect. To be specific, we design a feature adaptation module and plug it into all higher task-specific layers, including the classification layer.…”

Section: Domain-general Feature Learningmentioning

confidence: 99%

Domain Conditioned Adaptation Network

Liu

Lin

et al. 2020

AAAI

Self Cite

View full text Add to dashboard Cite

show abstract

“…Since the machine is in a healthy working state most of the time, the test data may contain only a few types of fault data. That is, the distribution of two domains is different and the label space of target domain is a subset of that of the source domain [ 20 , 21 , 22 ]. As many different health types as possible can be involved by training data through a long period of data accumulation, while it is difficult to guarantee the symmetry of health types in testing data and training data.…”

Section: Introductionmentioning

confidence: 99%

A Weighted Subdomain Adaptation Network for Partial Transfer Fault Diagnosis of Rotating Machinery

Jia

Wang

Zhang

et al. 2021

Entropy

View full text Add to dashboard Cite

Domain adaptation-based models for fault classification under variable working conditions have become a research focus in recent years. Previous domain adaptation approaches generally assume identical label spaces in the source and target domains, however, such an assumption may be no longer legitimate in a more realistic situation that requires adaptation from a larger and more diverse source domain to a smaller target domain with less number of fault classes. To address the above deficiencies, we propose a partial transfer fault diagnosis model based on a weighted subdomain adaptation network (WSAN) in this paper. Our method pays more attention to the local data distribution while aligning the global distribution. An auxiliary classifier is introduced to obtain the class-level weights of the source samples, so the network can avoid negative transfer caused by unique fault classes in the source domain. Furthermore, a weighted local maximum mean discrepancy (WLMMD) is proposed to capture the fine-grained transferable information and obtain sample-level weights. Finally, relevant distributions of domain-specific layer activations across different domains are aligned. Experimental results show that our method could assign appropriate weights to each source sample and realize efficient partial transfer fault diagnosis.

show abstract

“…Deep domain adaptation methods attempt to improve the learning of domain-invariant feature representations by em- * Corresponding Author bedding distribution matching modules into the network architectures. So far, a variety of distribution similarity measures have been incorporated into the network architecture to learn transferable representations [Long et al, 2017]. Meanwhile, a collection of adversarial learning based domain adaptation methods have been proposed, which aligns domain distributions via minimizing an approximate discrepancy in an adversarial training setting [Ganin and Lempitsky, 2014].…”

Section: Introductionmentioning

confidence: 99%

“…Nevertheless, in real-world applications, it is often nontrivial to find source domains with the same label spaces as target domains of interest. Instead, due to the availability of large-scale labelled datasets, such as ImageNet [Russakovsky et al, 2015], a more practical yet more challenging scenario is referred to as partial domain adaptation (PDA), which relaxes the constraint of shared label spaces. It enables knowledge transfer from source domains with more classes to target domains with fewer classes without any knowledge on the size and categories of the target classes.…”

Section: Introductionmentioning

confidence: 99%

Deep Reinforcement Learning Boosted Partial Domain Adaptation

Yang

et al. 2021

Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence

View full text Add to dashboard Cite

Domain adaptation is critical for learning transferable features that effectively reduce the distribution difference among domains. In the era of big data, the availability of large-scale labeled datasets motivates partial domain adaptation (PDA) which deals with adaptation from large source domains to small target domains with less number of classes. In the PDA setting, it is crucial to transfer relevant source samples and eliminate irrelevant ones to mitigate negative transfer. In this paper, we propose a deep reinforcement learning based source data selector for PDA, which is capable of eliminating less relevant source samples automatically to boost existing adaptation methods. It determines to either keep or discard the source instances based on their feature representations so that more effective knowledge transfer across domains can be achieved via filtering out irrelevant samples. As a general module, the proposed DRL-based data selector can be integrated into any existing domain adaptation or partial domain adaptation models. Extensive experiments on several benchmark datasets demonstrate the superiority of the proposed DRL-based data selector which leads to state-of-the-art performance for various PDA tasks.

show abstract

Deep Residual Correction Network for Partial Domain Adaptation

Cited by 132 publications

References 34 publications

Domain Conditioned Adaptation Network

Domain Conditioned Adaptation Network

A Weighted Subdomain Adaptation Network for Partial Transfer Fault Diagnosis of Rotating Machinery

Deep Reinforcement Learning Boosted Partial Domain Adaptation

Contact Info

Product

Resources

About