Classic GNNs are Strong Baselines: Reassessing GNNs for Node Classification

计算机科学 生物
作者
Yuankai Luo,Lei Shi,Xiao-Ming Wu
出处
期刊:Cornell University - arXiv
标识
DOI:10.48550/arxiv.2406.08993
摘要

Graph Transformers (GTs) have recently emerged as popular alternatives to traditional message-passing Graph Neural Networks (GNNs), due to their theoretically superior expressiveness and impressive performance reported on standard node classification benchmarks, often significantly outperforming GNNs. In this paper, we conduct a thorough empirical analysis to reevaluate the performance of three classic GNN models (GCN, GAT, and GraphSAGE) against GTs. Our findings suggest that the previously reported superiority of GTs may have been overstated due to suboptimal hyperparameter configurations in GNNs. Remarkably, with slight hyperparameter tuning, these classic GNN models achieve state-of-the-art performance, matching or even exceeding that of recent GTs across 17 out of the 18 diverse datasets examined. Additionally, we conduct detailed ablation studies to investigate the influence of various GNN configurations, such as normalization, dropout, residual connections, network depth, and jumping knowledge mode, on node classification performance. Our study aims to promote a higher standard of empirical rigor in the field of graph machine learning, encouraging more accurate comparisons and evaluations of model capabilities.
最长约 10秒,即可获得该文献文件

科研通智能强力驱动
Strongly Powered by AbleSci AI
科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
3秒前
科研通AI5应助小文采纳,获得10
3秒前
3秒前
5秒前
叶世玉发布了新的文献求助10
7秒前
tianwu发布了新的文献求助10
9秒前
9秒前
10秒前
10秒前
药小博完成签到,获得积分10
11秒前
科目三应助菜宝儿采纳,获得30
13秒前
CipherSage应助wjr采纳,获得10
15秒前
xgx984发布了新的文献求助10
15秒前
小文发布了新的文献求助10
15秒前
CipherSage应助风中的寻凝采纳,获得10
16秒前
Oracle应助ShengQ采纳,获得20
18秒前
小二郎应助豆子采纳,获得10
20秒前
海豚完成签到,获得积分10
20秒前
SYF发布了新的文献求助30
23秒前
清秀千兰完成签到,获得积分10
24秒前
27秒前
笑点低半仙给笑点低半仙的求助进行了留言
27秒前
大排量发布了新的文献求助10
27秒前
zmw完成签到,获得积分10
28秒前
小文完成签到,获得积分10
31秒前
海豚发布了新的文献求助10
31秒前
搜集达人应助xxn采纳,获得10
32秒前
yelis完成签到,获得积分10
34秒前
34秒前
威武从寒发布了新的文献求助20
34秒前
34秒前
xio完成签到,获得积分10
35秒前
852应助兴奋的蜡烛采纳,获得10
35秒前
研友_VZG7GZ应助英俊白莲采纳,获得30
35秒前
36秒前
安生完成签到,获得积分10
39秒前
lone623发布了新的文献求助10
39秒前
Ray发布了新的文献求助10
39秒前
乐乐应助大力水饺采纳,获得10
40秒前
41秒前
高分求助中
Worked Bone, Antler, Ivory, and Keratinous Materials 1000
Mass producing individuality 600
Algorithmic Mathematics in Machine Learning 500
Разработка метода ускоренного контроля качества электрохромных устройств 500
Getting Published in SSCI Journals: 200+ Questions and Answers for Absolute Beginners 300
Advances in Underwater Acoustics, Structural Acoustics, and Computational Methodologies 300
Limes XXIII Sonderband 4 / II Proceedings of the 23rd International Congress of Roman Frontier Studies Ingolstadt 2015 200
热门求助领域 (近24小时)
化学 材料科学 医学 生物 工程类 有机化学 物理 生物化学 纳米技术 计算机科学 化学工程 内科学 复合材料 物理化学 电极 遗传学 量子力学 基因 冶金 催化作用
热门帖子
关注 科研通微信公众号,转发送积分 3829369
求助须知:如何正确求助?哪些是违规求助? 3372030
关于积分的说明 10470309
捐赠科研通 3091581
什么是DOI,文献DOI怎么找? 1701245
邀请新用户注册赠送积分活动 818327
科研通“疑难数据库(出版商)”最低求助积分说明 770830