English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

In smart computing, the labels of training samples for a specific task arenot always abundant. However, the labels of samples in a relevant but differentdataset are available. As a result, researchers have relied on unsuperviseddomain adaptation to leverage the labels in a dataset (the source domain) toperform better classification in a different, unlabeled dataset (targetdomain). Existing non-generative adversarial solutions for UDA aim at achievingdomain confusion through adversarial training. The ideal scenario is thatperfect domain confusion is achieved, but this is not guaranteed to be true. Tofurther enforce domain confusion on top of the adversarial training, we proposea novel UDA algorithm, \textit{E-ADDA}, which uses both a novel variation ofthe Mahalanobis distance loss and an out-of-distribution detection subroutine.The Mahalanobis distance loss minimizes the distribution-wise distance betweenthe encoded target samples and the distribution of the source domain, thusenforcing additional domain confusion on top of adversarial training. Then, theOOD subroutine further eliminates samples on which the domain confusion isunsuccessful. We have performed extensive and comprehensive evaluations ofE-ADDA in the acoustic and computer vision modalities. In the acousticmodality, E-ADDA outperforms several state-of-the-art UDA algorithms by up to29.8%, measured in the f1 score. In the computer vision modality, theevaluation results suggest that we achieve new state-of-the-art performance onpopular UDA benchmarks such as Office-31 and Office-Home, outperforming thesecond best-performing algorithms by up to 17.9%.

E-ADDA: Unsupervised Adversarial Domain Adaptation Enhanced by a New Mahalanobis Distance Loss for Smart Computing