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

Supervised representation learning with deep networks tends to overfit thetraining classes and the generalization to novel classes is a challengingquestion. It is common to evaluate a learned embedding on held-out images ofthe same training classes. In real applications however, data comes from newsources and novel classes are likely to arise. We hypothesize thatincorporating unlabelled images of novel classes in the training set in asemi-supervised fashion would be beneficial for the efficient retrieval ofnovel-class images compared to a vanilla supervised representation. To verifythis hypothesis in a comprehensive way, we propose an original evaluationmethodology that varies the degree of novelty of novel classes by partitioningthe dataset category-wise either randomly, or semantically, i.e. by minimizingthe shared semantics between base and novel classes. This evaluation procedureallows to train a representation blindly to any novel-class labels and evaluatethe frozen representation on the retrieval of base or novel classes. We findthat a vanilla supervised representation falls short on the retrieval of novelclasses even more so when the semantics gap is higher. Semi-supervisedalgorithms allow to partially bridge this performance gap but there is stillmuch room for improvement.

How does the degree of novelty impacts semi-supervised representation learning for novel class retrieval?