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

In this work, we revisit the semi-supervised learning (SSL) problem from anew perspective of explicitly reducing empirical distribution mismatch betweenlabeled and unlabeled samples. Benefited from this new perspective, we firstpropose a new deep semi-supervised learning framework called Semi-supervisedLearning by Empirical Distribution Alignment (SLEDA), in which existingtechnologies from the domain adaptation community can be readily used toaddress the semi-supervised learning problem through reducing the empiricaldistribution distance between labeled and unlabeled data. Based on thisframework, we also develop a new theoretical generalization bound for theresearch community to better understand the semi-supervised learning problem,in which we show the generalization error of semi-supervised learning can beeffectively bounded by minimizing the training error on labeled data and theempirical distribution distance between labeled and unlabeled data. Buildingupon our new framework and the theoretical bound, we develop a simple andeffective deep semi-supervised learning method called Augmented DistributionAlignment Network (ADA-Net) by simultaneously adopting the well-establishedadversarial training strategy from the domain adaptation community and a simplesample interpolation strategy for data augmentation. Additionally, weincorporate both strategies in our ADA-Net into two exiting SSL methods tofurther improve their generalization capability, which indicates that our newframework provides a complementary solution for solving the SSL problem. Ourcomprehensive experimental results on two benchmark datasets SVHN and CIFAR-10for the semi-supervised image recognition task and another two benchmarkdatasets ModelNet40 and ShapeNet55 for the semi-supervised point cloudrecognition task demonstrate the effectiveness of our proposed framework forSSL.

Revisiting Deep Semi-supervised Learning: An Empirical Distribution Alignment Framework and Its Generalization Bound