We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Fea2Fea: Exploring Structural Feature Correlations via Graph Neural Networks

Abstract: Structural features are important features in a geometrical graph. Although there are some correlation analysis of features based on covariance, there is no relevant research on structural feature correlation analysis with graph neural networks. In this paper, we introuduce graph feature to feature (Fea2Fea) prediction pipelines in a low dimensional space to explore some preliminary results on structural feature correlation, which is based on graph neural network. The results show that there exists high correlation between some of the structural features. An irredundant feature combination with initial node features, which is filtered by graph neural network has improved its classification accuracy in some graph-based tasks. We compare differences between concatenation methods on connecting embeddings between features and show that the simplest is the best. We generalize on the synthetic geometric graphs and certify the results on prediction difficulty between structural features.
Comments: In proceddings of ECML-PKDD 2021 workshop
Subjects: Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Social and Information Networks (cs.SI)
Cite as: arXiv:2106.13061 [cs.LG]
  (or arXiv:2106.13061v4 [cs.LG] for this version)

Submission history

From: Jiaqing Xie [view email]
[v1] Thu, 24 Jun 2021 14:36:50 GMT (1776kb)
[v2] Thu, 1 Jul 2021 01:54:07 GMT (1776kb)
[v3] Sat, 17 Jul 2021 08:34:13 GMT (1776kb)
[v4] Thu, 9 Sep 2021 04:38:35 GMT (1773kb)

Link back to: arXiv, form interface, contact.