Advancing Imbalanced Domain Adaptation: Cluster-Level Discrepancy Minimization With a Comprehensive Benchmark

Yang, Jianfei; Yang, Jiangang; Wang, Shizheng; Cao, Shuxin; Zou, Han; Xie, Lihua

doi:10.1109/tcyb.2021.3093888

Cited by 20 publications

(8 citation statements)

References 44 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Domain adaptation is a type of semi-supervised learning that addresses domain shifts in transfer learning. It has been successfully implemented in object recognition 76 , 77 , 78 and text classification. 79 Multidisciplinary scenarios are common in WiFi sensing because CSI data are significantly dependent on the training environment.…”

Section: Resultsmentioning

confidence: 99%

SenseFi: A library and benchmark on deep-learning-empowered WiFi human sensing

Yang¹,

Chen²,

Zou³

et al. 2023

Patterns

Self Cite

View full text Add to dashboard Cite

Section: Resultsmentioning

confidence: 99%

SenseFi: A library and benchmark on deep-learning-empowered WiFi human sensing

Yang¹,

Chen²,

Zou³

et al. 2023

Patterns

Self Cite

View full text Add to dashboard Cite

“…Current UDA and VUDA methods aim to transfer knowledge from the source to the target domain given that both domains contain sufficient data, improving the transferability and robustness of models [55,62]. They could be generally divided into four categories: a) reconstruction-based methods [14,61], where domain-invariant features are obtained by encoders trained with data-reconstruction objectives; b) adversarial-based methods [5,56,8], where feature generators obtain domain-invariant features leveraging domain discriminators, trained jointly in an adversarial manner [16,10]; c) semantic-based methods [63,58], which exploit the shared semantics across domains such that domaininvariant features are obtained; and d) discrepancy-based methods [32,67], which mitigate domain shifts by applying metric learning, minimizing metrics such as MMD [24] and CORAL [37]. With the introduction of cross-domain video datasets such as Daily-DA [59] and Sports-DA [59], there has been a significant increase in research interest for VUDA [8,27,6].…”

Section: Related Workmentioning

confidence: 99%

Augmenting and Aligning Snippets for Few-Shot Video Domain Adaptation

Xu¹,

Yang²,

Zhou³

et al. 2023

Preprint

View full text Add to dashboard Cite

For video models to be transferred and applied seamlessly across video tasks in varied environments, Video Unsupervised Domain Adaptation (VUDA) has been introduced to improve the robustness and transferability of video models. However, current VUDA methods rely on a vast amount of high-quality unlabeled target data, which may not be available in real-world cases. We thus consider a more realistic Few-Shot Video-based Domain Adaptation (FSVDA) scenario where we adapt video models with only a few target video samples. While a few methods have touched upon Few-Shot Domain Adaptation (FSDA) in images and in FSVDA, they rely primarily on spatial augmentation for target domain expansion with alignment performed statistically at the instance level. However, videos contain more knowledge in terms of rich temporal and semantic information, which should be fully considered while augmenting target domains and performing alignment in FSVDA. We propose a novel SSA 2 lign to address FSVDA at the snippet level, where the target domain is expanded through a simple snippet-level augmentation followed by the attentive alignment of snippets both semantically and statistically, where semantic alignment of snippets is conducted through multiple perspectives. Empirical results demonstrate state-of-the-art performance of SSA 2 lign across multiple cross-domain action recognition benchmarks.

show abstract

“…Domain adversarial learning further employs a domain discriminator to achieve the same goal [20,46,26,85,67,77,76,74] and achieves remarkable results. Other effective techniques for UDA include entropy minimization [21,15,60], contrastive learning [30,27], domain normalization [71,8], semantic alignment [38,72,16,75], meta-learning [44], self-supervision [57], curriculum learning [80] and self-training [10,86,58]. Despite their effectiveness, they require the access to the source domain data and therefore invoke privacy and portability concerns.…”

Section: Related Workmentioning

confidence: 99%

Divide to Adapt: Mitigating Confirmation Bias for Domain Adaptation of Black-Box Predictors

Yang¹,

Peng²,

Wang³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Domain Adaptation of Black-box Predictors (DABP) aims to learn a model on an unlabeled target domain supervised by a black-box predictor trained on a source domain. It does not require access to both the source-domain data and the predictor parameters, thus addressing the data privacy and portability issues of standard domain adaptation. Existing DABP approaches mostly rely on model distillation from the black-box predictor, i.e., training the model with its noisy target-domain predictions, which however inevitably introduces the confirmation bias accumulated from the prediction noises. To mitigate such bias, we propose a new method, named BETA, to incorporate knowledge distillation and noisy label learning into one coherent framework. This is enabled by a new divide-to-adapt strategy. BETA divides the target domain into an easy-to-adapt subdomain with less noise and a hard-to-adapt subdomain. Then it deploys mutually-teaching twin networks to filter the predictor errors for each other and improve them progressively, from the easy to hard subdomains. As such, BETA effectively purifies the noisy labels and reduces error accumulation. We theoretically show that the target error of BETA is minimized by decreasing the noise ratio of the subdomains. Extensive experiments demonstrate BETA outperforms existing methods on all DABP benchmarks, and is even comparable with the standard domain adaptation methods that use the source-domain data.

show abstract

Advancing Imbalanced Domain Adaptation: Cluster-Level Discrepancy Minimization With a Comprehensive Benchmark

Cited by 20 publications

References 44 publications

SenseFi: A library and benchmark on deep-learning-empowered WiFi human sensing

SenseFi: A library and benchmark on deep-learning-empowered WiFi human sensing

Augmenting and Aligning Snippets for Few-Shot Video Domain Adaptation

Divide to Adapt: Mitigating Confirmation Bias for Domain Adaptation of Black-Box Predictors

Contact Info

Product

Resources

About