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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Heterogeneous information network (HIN) embedding, aiming to map thestructure and semantic information in a HIN to distributed representations, hasdrawn considerable research attention. Graph neural networks for HIN embeddingstypically adopt a hierarchical attention (including node-level andmeta-path-level attentions) to capture the information from meta-path-basedneighbors. However, this complicated attention structure often cannot achievethe function of selecting meta-paths due to severe overfitting. Moreover, whenpropagating information, these methods do not distinguish direct (one-hop)meta-paths from indirect (multi-hop) ones. But from the perspective of networkscience, direct relationships are often believed to be more essential, whichcan only be used to model direct information propagation. To address theselimitations, we propose a novel neural network method via implicitly utilizingattention and meta-paths, which can relieve the severe overfitting brought bythe current over-parameterized attention mechanisms on HIN. We first use themulti-layer graph convolutional network (GCN) framework, which performs adiscriminative aggregation at each layer, along with stacking the informationpropagation of direct linked meta-paths layer-by-layer, realizing the functionof attentions for selecting meta-paths in an indirect way. We then give aneffective relaxation and improvement via introducing a new propagationoperation which can be separated from aggregation. That is, we first model thewhole propagation process with well-defined probabilistic diffusion dynamics,and then introduce a random graph-based constraint which allows it to reducenoise with the increase of layers. Extensive experiments demonstrate thesuperiority of the new approach over state-of-the-art methods.

GCN for HIN via Implicit Utilization of Attention and Meta-paths