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

Designing robust algorithms capable of training accurate neural networks onuncurated datasets from the web has been the subject of much research as itreduces the need for time consuming human labor. The focus of many previousresearch contributions has been on the detection of different types of labelnoise; however, this paper proposes to improve the correction accuracy of noisysamples once they have been detected. In many state-of-the-art contributions, atwo phase approach is adopted where the noisy samples are detected beforeguessing a corrected pseudo-label in a semi-supervised fashion. The guessedpseudo-labels are then used in the supervised objective without ensuring thatthe label guess is likely to be correct. This can lead to confirmation bias,which reduces the noise robustness. Here we propose the pseudo-loss, a simplemetric that we find to be strongly correlated with pseudo-label correctness onnoisy samples. Using the pseudo-loss, we dynamically down weightunder-confident pseudo-labels throughout training to avoid confirmation biasand improve the network accuracy. We additionally propose to use a confidenceguided contrastive objective that learns robust representation on aninterpolated objective between class bound (supervised) for confidentlycorrected samples and unsupervised representation for under-confident labelcorrections. Experiments demonstrate the state-of-the-art performance of ourPseudo-Loss Selection (PLS) algorithm on a variety of benchmark datasetsincluding curated data synthetically corrupted with in-distribution andout-of-distribution noise, and two real world web noise datasets. Ourexperiments are fully reproducible github.com/PaulAlbert31/SNCF

Is your noise correction noisy? PLS: Robustness to label noise with two stage detection