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 Plus annual

Presents the access of premium content as premium feature

Premium Content

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 Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Unsupervised domain adaptation reduces the reliance on data annotation indeep learning by adapting knowledge from a source to a target domain. Forprivacy and efficiency concerns, source-free domain adaptation extendsunsupervised domain adaptation by adapting a pre-trained source model to anunlabeled target domain without accessing the source data. However, mostexisting source-free domain adaptation methods to date focus on thetransductive setting, where the target training set is also the testing set. Inthis paper, we address source-free domain adaptation in the more realisticinductive setting, where the target training and testing sets are mutuallyexclusive. We propose a new semi-supervised fine-tuning method named DualMoving Average Pseudo-Labeling (DMAPL) for source-free inductive domainadaptation. We first split the unlabeled training set in the target domain intoa pseudo-labeled confident subset and an unlabeled less-confident subsetaccording to the prediction confidence scores from the pre-trained sourcemodel. Then we propose a soft-label moving-average updating strategy for theunlabeled subset based on a moving-average prototypical classifier, whichgradually adapts the source model towards the target domain. Experiments showthat our proposed method achieves state-of-the-art performance and outperformsprevious methods by large margins.

Dual Moving Average Pseudo-Labeling for Source-Free Inductive Domain Adaptation