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

Graph neural networks (GNN) suffer from severe inefficiency. It is mainlycaused by the exponential growth of node dependency with the increase oflayers. It extremely limits the application of stochastic optimizationalgorithms so that the training of GNN is usually time-consuming. To addressthis problem, we propose to decouple a multi-layer GNN as multiple simplemodules for more efficient training, which is comprised of classical forwardtraining (FT)and designed backward training (BT). Under the proposed framework,each module can be trained efficiently in FT by stochastic algorithms withoutdistortion of graph information owing to its simplicity. To avoid the onlyunidirectional information delivery of FT and sufficiently train shallowmodules with the deeper ones, we develop a backward training mechanism thatmakes the former modules perceive the latter modules. The backward trainingintroduces the reversed information delivery into the decoupled modules as wellas the forward information delivery. To investigate how the decoupling andgreedy training affect the representational capacity, we theoretically provethat the error produced by linear modules will not accumulate on unsupervisedtasks in most cases. The theoretical and experimental results show that theproposed framework is highly efficient with reasonable performance.

Decouple Graph Neural Networks: Train Multiple Simple GNNs Simultaneously Instead of One