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

Due to inappropriate sample selection and limited training data, adistribution shift often exists between the training and test sets. This shiftcan adversely affect the test performance of Graph Neural Networks (GNNs).Existing approaches mitigate this issue by either enhancing the robustness ofGNNs to distribution shift or reducing the shift itself. However, bothapproaches necessitate retraining the model, which becomes unfeasible when themodel structure and parameters are inaccessible. To address this challenge, wepropose FR-GNN, a general framework for GNNs to conduct feature reconstruction.FRGNN constructs a mapping relationship between the output and input of awell-trained GNN to obtain class representative embeddings and then uses theseembeddings to reconstruct the features of labeled nodes. These reconstructedfeatures are then incorporated into the message passing mechanism of GNNs toinfluence the predictions of unlabeled nodes at test time. Notably, thereconstructed node features can be directly utilized for testing thewell-trained model, effectively reducing the distribution shift and leading toimproved test performance. This remarkable achievement is attained without anymodifications to the model structure or parameters. We provide theoreticalguarantees for the effectiveness of our framework. Furthermore, we conductcomprehensive experiments on various public datasets. The experimental resultsdemonstrate the superior performance of FRGNN in comparison to multiplecategories of baseline methods.

FRGNN: Mitigating the Impact of Distribution Shift on Graph Neural Networks via Test-Time Feature Reconstruction