{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/neighborhood-homophily-guided-graph-1","title":"Neighborhood Homophily-Guided Graph Convolutional Network","arxiv_id":null,"date":"2023-10-21","proceeding":"CIKM 2023 10","authors":["Shengbo Gong","Jiajun Zhou","Chenxuan Xie","Qi Xuan"],"abstract":"Graph neural networks (GNNs) have been proved powerful in\r\ngraph-oriented tasks. However, many real-world graphs are heterophilous, challenging the homophily assumption of classical\r\nGNNs. To solve the universality problem, many studies deepen\r\nnetworks or concatenate intermediate representations, which does\r\nnot inherently change neighbor aggregation and introduces noise.\r\nRecent studies propose new metrics to characterize the homophily,\r\nbut rarely consider the correlation of the proposed metrics and\r\nmodels. In this paper, we first design a new metric, Neighborhood\r\nHomophily (NH), to measure the label complexity or purity in\r\nnode neighborhoods. Furthermore, we incorporate the metric into\r\nthe classical graph convolutional network (GCN) architecture and\r\npropose Neighborhood Homophily-based Graph Convolutional\r\nNetwork (NHGCN). In this framework, neighbors are grouped by\r\nestimated NH values and aggregated from different channels, and\r\nthe resulting node predictions are then used in turn to estimate\r\nand update NH values. The two processes of metric estimation and\r\nmodel inference are alternately optimized to achieve better node\r\nclassification. NHGCN achieves top overall performance on both\r\nhomophilous and heterophilous benchmarks, with an improvement\r\nof up to 7.4% compared to the current SOTA methods","url_abs":"https://dl.acm.org/doi/10.1145/3583780.3615195","url_pdf":"https://arxiv.org/pdf/2301.09851","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"neighborhood-homophily-guided-graph-1","repo_url":"https://github.com/rockcor/NHGCN","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"node-classification","task_name":"Node Classification"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/node-classification-on-film-60-20-20-random","task":"Node Classification","dataset":"Film (60%/20%/20% random splits)","model":"NHGCN","rank_in_archive_order":2,"of":37,"metrics":{"1:1 Accuracy":"43.94 ± 1.14"},"uses_additional_data":false},{"leaderboard":"/sota/node-classification-on-pubmed-60-20-20-random","task":"Node Classification","dataset":"PubMed (60%/20%/20% random splits)","model":"NHGCN","rank_in_archive_order":2,"of":37,"metrics":{"1:1 Accuracy":"91.56 ± 0.50"},"uses_additional_data":false}],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}