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/

Zendy Plus

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

Zendy Tools

Zendy Open

Unsupervised Contrastive learning has gained prominence in fields such asvision, and biology, leveraging predefined positive/negative samples forrepresentation learning. Data augmentation, categorized into hand-designed andmodel-based methods, has been identified as a crucial component for enhancingcontrastive learning. However, hand-designed methods require human expertise indomain-specific data while sometimes distorting the meaning of the data. Incontrast, generative model-based approaches usually require supervised orlarge-scale external data, which has become a bottleneck constraining modeltraining in many domains. To address the problems presented above, this paperproposes DiffAug, a novel unsupervised contrastive learning technique withdiffusion mode-based positive data generation. DiffAug consists of a semanticencoder and a conditional diffusion model; the conditional diffusion modelgenerates new positive samples conditioned on the semantic encoding to servethe training of unsupervised contrast learning. With the help of iterativetraining of the semantic encoder and diffusion model, DiffAug improves therepresentation ability in an uninterrupted and unsupervised manner.Experimental evaluations show that DiffAug outperforms hand-designed and SOTAmodel-based augmentation methods on DNA sequence, visual, and bio-featuredatasets. The code for review is released at\url{https://github.com/zangzelin/code_diffaug}.

DiffAug: Enhance Unsupervised Contrastive Learning with Domain-Knowledge-Free Diffusion-based Data Augmentation