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

Despite the large progress in supervised learning with neural networks, thereare significant challenges in obtaining high-quality, large-scale andaccurately labelled datasets. In such a context, how to learn in the presenceof noisy labels has received more and more attention. As a relatively complexproblem, in order to achieve good results, current approaches often integratecomponents from several fields, such as supervised learning, semi-supervisedlearning, transfer learning and resulting in complicated methods. Furthermore,they often make multiple assumptions about the type of noise of the data. Thisaffects the model robustness and limits its performance under different noiseconditions. In this paper, we consider a novel problem setting, Learning withUnknown Label Noise}(LULN), that is, learning when both the degree and the typeof noise are unknown. Under this setting, unlike previous methods that oftenintroduce multiple assumptions and lead to complex solutions, we propose asimple, efficient and robust framework named Sample Selection andRelabelling(SSR), that with a minimal number of hyperparameters achieves SOTAresults in various conditions. At the heart of our method is a sample selectionand relabelling mechanism based on a non-parametric KNN classifier~(NPK) $g_q$and a parametric model classifier~(PMC) $g_p$, respectively, to select theclean samples and gradually relabel the noisy samples. Without bells andwhistles, such as model co-training, self-supervised pre-training andsemi-supervised learning, and with robustness concerning the settings of itsfew hyper-parameters, our method significantly surpasses previous methods onboth CIFAR10/CIFAR100 with synthetic noise and real-world noisy datasets suchas WebVision, Clothing1M and ANIMAL-10N. Code is available athttps://github.com/MrChenFeng/SSR_BMVC2022.

SSR: An Efficient and Robust Framework for Learning with Unknown Label Noise