TrustSGCN: Learning Trustworthiness on Edge Signs for Effective Signed Graph Convolutional Networs
Conclusions
Conclusions
In this work, we explored the limitations of existing GCN-based SNE methods, which arise when relying blindly on the rules of bal- ance theory. To address the limitations, we proposed a novel SNE method, TrustSGCN, which learns trustworthiness on edge signs for signed graph convolutional networks. Via experiments using 4 real-world datasets, we demonstrated that TrustSGCN consistently outperforms 5 GCN-based SNE methods. To the best of our knowl- edge, we are the first who designs a GCN architecture that can consider the trustworthiness on edge signs, being able to encourage follow-up studies on trustworthy SNE research [34].
In this work, we explored the limitations of existing GCN-based SNE methods, which arise when relying blindly on the rules of bal- ance theory. To address the limitations, we proposed a novel SNE method, TrustSGCN, which learns trustworthiness on edge signs for signed graph convolutional networks. Via experiments using 4 real-world datasets, we demonstrated that TrustSGCN consistently outperforms 5 GCN-based SNE methods. To the best of our knowl- edge, we are the first who designs a GCN architecture that can consider the trustworthiness on edge signs, being able to encourage follow-up studies on trustworthy SNE research [34].