Multi-head collaborative learning for graph neural networks

Haiyun Xu

doi:10.1016/j.neucom.2022.05.027

Multi-head collaborative learning for graph neural networks

2022-08-01

6

PRE

AI

张绍杰 (Shaojie Zhang) *

DOI：10.1016/j.neucom.2022.05.027

Original

Original request for help

Save

Abstract

En 中文

Recently, graph neural networks (GNNs) have been widely used for graph-structured data representation. However, existing GNNs generally either employ one classifier head which usually makes the model fall into local minima or multiple classifier heads independently which lack of exploiting the correlation of them. To overcome this issue, in this paper, we propose to employ Multi-head Collaborative Learning for Graph Neural Networks (MCL-GNNs). The MCL-GNNs provide a collaborative learning process via multiple classifiers to complement each other's information. Experimental results on several benchmark data sets show that our method has better generality and stability than other comparison methods. (c) 2022 Elsevier B.V. All rights reserved.

Keywords:

Collaborative learning

Classification

Non-Euclidean data

Graph network

Multi-head collaborative learning for graph neural networks

Abstract

Journal

Organization