A decoupled graph neural network with dropping repeated nodes

Huadong Li; Yan Yang

doi:10.1117/12.3011648

7 December 2023 A decoupled graph neural network with dropping repeated nodes

Huadong Li, Yan Yang

Proceedings Volume 12941, International Conference on Algorithms, High Performance Computing, and Artificial Intelligence (AHPCAI 2023); 129415H (2023) https://doi.org/10.1117/12.3011648
Event: Third International Conference on Algorithms, High Performance Computing, and Artificial Intelligence (AHPCAI 203), 2023, Yinchuan, China

Abstract

Graph Neural Networks (GNNs) and their variants have significant success in graph representation learning tasks. However, existing GNNs still need to overcome the challenges of over-smoothing and over-fitting. To address these issues, we propose a new decoupled graph neural network with DropRE sparsification named DropRE-GCN. Specifically, we first design a graph sparsification method called Drop Repetition (DropRE), which avoids over-smoothing in GNNs by dropping repetitive edges and making node connections sparser in graph. Different from the traditional random drop method in that it deterministically drops the nodes aggregated in the previous layer in each round of training to remove the randomness in the conventional drop method. In addition, based on the decoupled graph neural network framework, we pre-compute the process of multi-hop neighborhood aggregation to alleviate the overall computation. Finally, DropRE-GCN balances each node's local and global information by a node-level adaptive method. Extensive experimentation on the dataset indicates that DropRE-GCN performs better than other competing baseline models, alleviating the over-smoothing and over-fitting problems.

(2023) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Huadong Li and Yan Yang "A decoupled graph neural network with dropping repeated nodes", Proc. SPIE 12941, International Conference on Algorithms, High Performance Computing, and Artificial Intelligence (AHPCAI 2023), 129415H (7 December 2023); https://doi.org/10.1117/12.3011648

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