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

Class labels used for machine learning are relatable to each other, withcertain class labels being more similar to each other than others (e.g. imagesof cats and dogs are more similar to each other than those of cats and cars).Such similarity among classes is often the cause of poor model performance dueto the models confusing between them. Current labeling techniques fail toexplicitly capture such similarity information. In this paper, we insteadexploit the similarity between classes by capturing the similarity informationwith our novel confidence labels. Confidence labels are probabilistic labelsdenoting the likelihood of similarity, or confusability, between the classes.Often even after models are trained to differentiate between classes in thefeature space, the similar classes' latent space still remains clustered. Weview this type of clustering as valuable information and exploit it with ournovel projective loss functions. Our projective loss functions are designed towork with confidence labels with an ability to relax the loss penalty forerrors that confuse similar classes. We use our approach to train neuralnetworks with noisy labels, as we believe noisy labels are partly a result ofconfusability arising from class similarity. We show improved performancecompared to the use of standard loss functions. We conduct a detailed analysisusing the CIFAR-10 dataset and show our proposed methods' applicability tolarger datasets, such as ImageNet and Food-101N.

Exploiting Class Similarity for Machine Learning with Confidence Labels and Projective Loss Functions