Masg: masked auto-encoder for signed graph

Dengdi Sun; Zhenyu Wang; Bin Luo; Zhuanlian Ding

doi:10.1117/12.3032897

5 July 2024 Masg: masked auto-encoder for signed graph

Dengdi Sun, Zhenyu Wang, Bin Luo, Zhuanlian Ding

Proceedings Volume 13184, Third International Conference on Electronic Information Engineering and Data Processing (EIEDP 2024); 131844U (2024) https://doi.org/10.1117/12.3032897
Event: 3rd International Conference on Electronic Information Engineering and Data Processing (EIEDP 2024), 2024, Kuala Lumpur, Malaysia

Abstract

The graph neural network (GNN) is tailored for handling graph-structured data, efficiently capturing its topological structure and learning low-dimensional representations. Nonetheless, unsigned GNNs only work on unsigned networks, which have only positive edges. In contrast, social networks often exhibit a signed network structure, where edges can be either positive or negative. In social networks, there is a signed network in which edges have positive or negative signs. In this paper, we introduce the masking mechanism into signed networks and design a masked graph autoencoder suitable for signed networks. We randomly mask both positively and negatively weighted connections at a fixed ratio, incorporate the masked network into a signed graph convolution model for node representation learning, and subsequently train a decoder using logistic regression to reconstruct edge information for predicting the signs of the masked edges. Thorough empirical evaluations on four widely recognized social datasets demonstrate the superiority of our model, surpassing existing cutting-edge signed network models in terms of performance.

(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Dengdi Sun, Zhenyu Wang, Bin Luo, and Zhuanlian Ding "Masg: masked auto-encoder for signed graph", Proc. SPIE 13184, Third International Conference on Electronic Information Engineering and Data Processing (EIEDP 2024), 131844U (5 July 2024); https://doi.org/10.1117/12.3032897

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available