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

Generalized zero shot learning (GZSL) is still a technical challenge of deeplearning as it has to recognize both source and target classes without datafrom target classes. To preserve the semantic relation between source andtarget classes when only trained with data from source classes, we address thequantification of the knowledge transfer and semantic relation from aninformation-theoretic viewpoint. To this end, we follow the prototypical modeland format the variables of concern as a probability vector. Leveraging on theproposed probability vector representation, the information measurement such asmutual information and entropy, can be effectively evaluated with simple closedforms. We discuss the choice of common embedding space and distance functionwhen using the prototypical model. Then We propose three information-theoreticloss functions for deterministic GZSL model: a mutual information loss tobridge seen data and target classes; an uncertainty-aware entropy constraintloss to prevent overfitting when using seen data to learn the embedding oftarget classes; a semantic preserving cross entropy loss to preserve thesemantic relation when mapping the semantic representations to the commonspace. Simulation shows that, as a deterministic model, our proposed methodobtains state of the art results on GZSL benchmark datasets. We achieve 21%-64%improvements over the baseline model -- deep calibration network (DCN) and forthe first time demonstrate a deterministic model can perform as well asgenerative ones. Moreover, our proposed model is compatible with generativemodels. Simulation studies show that by incorporating with f-CLSWGAN, we obtaincomparable results compared with advanced generative models.

Prototypical Model with Novel Information-theoretic Loss Function for Generalized Zero Shot Learning