针对嗜异图形学习的参数化拉普拉奇灵活扩散范围

Qincheng Lu, Jiaqi Zhu, Sitao Luan, Xiao-Wen Chang
{"title":"针对嗜异图形学习的参数化拉普拉奇灵活扩散范围","authors":"Qincheng Lu, Jiaqi Zhu, Sitao Luan, Xiao-Wen Chang","doi":"arxiv-2409.09888","DOIUrl":null,"url":null,"abstract":"The ability of Graph Neural Networks (GNNs) to capture long-range and global\ntopology information is limited by the scope of conventional graph Laplacian,\nleading to unsatisfactory performance on some datasets, particularly on\nheterophilic graphs. To address this limitation, we propose a new class of\nparameterized Laplacian matrices, which provably offers more flexibility in\ncontrolling the diffusion distance between nodes than the conventional graph\nLaplacian, allowing long-range information to be adaptively captured through\ndiffusion on graph. Specifically, we first prove that the diffusion distance\nand spectral distance on graph have an order-preserving relationship. With this\nresult, we demonstrate that the parameterized Laplacian can accelerate the\ndiffusion of long-range information, and the parameters in the Laplacian enable\nflexibility of the diffusion scopes. Based on the theoretical results, we\npropose topology-guided rewiring mechanism to capture helpful long-range\nneighborhood information for heterophilic graphs. With this mechanism and the\nnew Laplacian, we propose two GNNs with flexible diffusion scopes: namely the\nParameterized Diffusion based Graph Convolutional Networks (PD-GCN) and Graph\nAttention Networks (PD-GAT). Synthetic experiments reveal the high correlations\nbetween the parameters of the new Laplacian and the performance of\nparameterized GNNs under various graph homophily levels, which verifies that\nour new proposed GNNs indeed have the ability to adjust the parameters to\nadaptively capture the global information for different levels of heterophilic\ngraphs. They also outperform the state-of-the-art (SOTA) models on 6 out of 7\nreal-world benchmark datasets, which further confirms their superiority.","PeriodicalId":501032,"journal":{"name":"arXiv - CS - Social and Information Networks","volume":null,"pages":null},"PeriodicalIF":0.0000,"publicationDate":"2024-09-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Flexible Diffusion Scopes with Parameterized Laplacian for Heterophilic Graph Learning\",\"authors\":\"Qincheng Lu, Jiaqi Zhu, Sitao Luan, Xiao-Wen Chang\",\"doi\":\"arxiv-2409.09888\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"The ability of Graph Neural Networks (GNNs) to capture long-range and global\\ntopology information is limited by the scope of conventional graph Laplacian,\\nleading to unsatisfactory performance on some datasets, particularly on\\nheterophilic graphs. To address this limitation, we propose a new class of\\nparameterized Laplacian matrices, which provably offers more flexibility in\\ncontrolling the diffusion distance between nodes than the conventional graph\\nLaplacian, allowing long-range information to be adaptively captured through\\ndiffusion on graph. Specifically, we first prove that the diffusion distance\\nand spectral distance on graph have an order-preserving relationship. With this\\nresult, we demonstrate that the parameterized Laplacian can accelerate the\\ndiffusion of long-range information, and the parameters in the Laplacian enable\\nflexibility of the diffusion scopes. Based on the theoretical results, we\\npropose topology-guided rewiring mechanism to capture helpful long-range\\nneighborhood information for heterophilic graphs. With this mechanism and the\\nnew Laplacian, we propose two GNNs with flexible diffusion scopes: namely the\\nParameterized Diffusion based Graph Convolutional Networks (PD-GCN) and Graph\\nAttention Networks (PD-GAT). Synthetic experiments reveal the high correlations\\nbetween the parameters of the new Laplacian and the performance of\\nparameterized GNNs under various graph homophily levels, which verifies that\\nour new proposed GNNs indeed have the ability to adjust the parameters to\\nadaptively capture the global information for different levels of heterophilic\\ngraphs. They also outperform the state-of-the-art (SOTA) models on 6 out of 7\\nreal-world benchmark datasets, which further confirms their superiority.\",\"PeriodicalId\":501032,\"journal\":{\"name\":\"arXiv - CS - Social and Information Networks\",\"volume\":null,\"pages\":null},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-09-15\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"arXiv - CS - Social and Information Networks\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/arxiv-2409.09888\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - CS - Social and Information Networks","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2409.09888","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0

摘要

图神经网络(GNN)捕捉远距离和全局拓扑信息的能力受到传统图拉普拉斯矩阵范围的限制,导致其在某些数据集上的性能不尽如人意,尤其是在嗜异图上。针对这一局限性,我们提出了一类新的参数化拉普拉斯矩阵,与传统图拉普拉斯矩阵相比,该矩阵在控制节点间扩散距离方面具有更大的灵活性,允许通过图上的扩散自适应地捕捉长程信息。具体来说,我们首先证明了图上的扩散距离和光谱距离具有保序关系。根据这一结果,我们证明了参数化的拉普拉斯可以加速长程信息的扩散,并且拉普拉斯中的参数可以实现扩散范围的灵活性。在理论结果的基础上,我们提出了拓扑引导的重布线机制,以捕捉异嗜图中有用的长邻域信息。利用这种机制和新的拉普拉斯函数,我们提出了两种具有灵活扩散范围的 GNN:即基于参数化扩散的图卷积网络(Parameterized Diffusion based Graph Convolutional Networks,PD-GCN)和图注意力网络(GraphAttention Networks,PD-GAT)。合成实验显示,在不同的图同亲程度下,新拉普拉斯参数与参数化 GNN 的性能之间存在高度相关性,这验证了我们提出的新 GNN 确实有能力调整参数,以适应性地捕捉不同异亲程度图的全局信息。在 7 个真实世界基准数据集中的 6 个数据集上,它们的表现也优于最先进的(SOTA)模型,这进一步证实了它们的优越性。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
Flexible Diffusion Scopes with Parameterized Laplacian for Heterophilic Graph Learning
The ability of Graph Neural Networks (GNNs) to capture long-range and global topology information is limited by the scope of conventional graph Laplacian, leading to unsatisfactory performance on some datasets, particularly on heterophilic graphs. To address this limitation, we propose a new class of parameterized Laplacian matrices, which provably offers more flexibility in controlling the diffusion distance between nodes than the conventional graph Laplacian, allowing long-range information to be adaptively captured through diffusion on graph. Specifically, we first prove that the diffusion distance and spectral distance on graph have an order-preserving relationship. With this result, we demonstrate that the parameterized Laplacian can accelerate the diffusion of long-range information, and the parameters in the Laplacian enable flexibility of the diffusion scopes. Based on the theoretical results, we propose topology-guided rewiring mechanism to capture helpful long-range neighborhood information for heterophilic graphs. With this mechanism and the new Laplacian, we propose two GNNs with flexible diffusion scopes: namely the Parameterized Diffusion based Graph Convolutional Networks (PD-GCN) and Graph Attention Networks (PD-GAT). Synthetic experiments reveal the high correlations between the parameters of the new Laplacian and the performance of parameterized GNNs under various graph homophily levels, which verifies that our new proposed GNNs indeed have the ability to adjust the parameters to adaptively capture the global information for different levels of heterophilic graphs. They also outperform the state-of-the-art (SOTA) models on 6 out of 7 real-world benchmark datasets, which further confirms their superiority.
求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
自引率
0.00%
发文量
0
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:481959085
Book学术官方微信