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

Modern deep neural networks can easily overfit to biased training datacontaining corrupted labels or class imbalance. Sample re-weighting methods arepopularly used to alleviate this data bias issue. Most current methods,however, require to manually pre-specify the weighting schemes as well as theiradditional hyper-parameters relying on the characteristics of the investigatedproblem and training data. This makes them fairly hard to be generally appliedin practical scenarios, due to their significant complexities and inter-classvariations of data bias situations. To address this issue, we propose ameta-model capable of adaptively learning an explicit weighting scheme directlyfrom data. Specifically, by seeing each training class as a separate learningtask, our method aims to extract an explicit weighting function with sampleloss and task/class feature as input, and sample weight as output, expecting toimpose adaptively varying weighting schemes to different sample classes basedon their own intrinsic bias characteristics. Synthetic and real dataexperiments substantiate the capability of our method on achieving properweighting schemes in various data bias cases, like the class imbalance,feature-independent and dependent label noise scenarios, and more complicatedbias scenarios beyond conventional cases. Besides, the task-transferability ofthe learned weighting scheme is also substantiated, by readily deploying theweighting function learned on relatively smaller-scale CIFAR-10 dataset on muchlarger-scale full WebVision dataset. A performance gain can be readily achievedcompared with previous SOAT ones without additional hyper-parameter tuning andmeta gradient descent step. The general availability of our method for multiplerobust deep learning issues, including partial-label learning, semi-supervisedlearning and selective classification, has also been validated.

CMW-Net: Learning a Class-Aware Sample Weighting Mapping for Robust Deep Learning