{"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/graphmix-regularized-training-of-graph-neural","title":"GraphMix: Improved Training of GNNs for Semi-Supervised Learning","arxiv_id":"1909.11715","date":"2019-09-25","proceeding":null,"authors":["Vikas Verma","Meng Qu","Kenji Kawaguchi","Alex Lamb","Yoshua Bengio","Juho Kannala","Jian Tang"],"abstract":"We present GraphMix, a regularization method for Graph Neural Network based semi-supervised object classification, whereby we propose to train a fully-connected network jointly with the graph neural network via parameter sharing and interpolation-based regularization. Further, we provide a theoretical analysis of how GraphMix improves the generalization bounds of the underlying graph neural network, without making any assumptions about the \"aggregation\" layer or the depth of the graph neural networks. We experimentally validate this analysis by applying GraphMix to various architectures such as Graph Convolutional Networks, Graph Attention Networks and Graph-U-Net. Despite its simplicity, we demonstrate that GraphMix can consistently improve or closely match state-of-the-art performance using even simpler architectures such as Graph Convolutional Networks, across three established graph benchmarks: Cora, Citeseer and Pubmed citation network datasets, as well as three newly proposed datasets: Cora-Full, Co-author-CS and Co-author-Physics.","url_abs":"https://arxiv.org/abs/1909.11715v3","url_pdf":"https://arxiv.org/pdf/1909.11715v3.pdf","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":"graphmix-regularized-training-of-graph-neural","repo_url":"https://github.com/vikasverma1077/GraphMix","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":{"status":"ok"}}],"tasks":[{"task_slug":"generalization-bounds","task_name":"Generalization Bounds"},{"task_slug":"graph-attention","task_name":"Graph Attention"},{"task_slug":"graph-neural-network","task_name":"Graph Neural Network"},{"task_slug":"node-classification","task_name":"Node Classification"}],"methods":[{"method_slug":"graph-convolutional-networks","method_name":"Graph Convolutional Networks"},{"method_slug":"graph-neural-network","method_name":"Graph Neural Network"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/node-classification-on-bitcoin-alpha","task":"Node Classification","dataset":"Bitcoin-Alpha","model":"GraphMix (GCN)","rank_in_archive_order":1,"of":1,"metrics":{"F1-score":"0.6534"},"uses_additional_data":false},{"leaderboard":"/sota/node-classification-on-bitcoin-otc","task":"Node Classification","dataset":"Bitcoin-OTC","model":"GraphMix (GCN)","rank_in_archive_order":1,"of":1,"metrics":{"F1-score":"0.6635"},"uses_additional_data":false},{"leaderboard":"/sota/node-classification-on-citeseer-with-public","task":"Node Classification","dataset":"CiteSeer with Public Split: fixed 20 nodes per class","model":"GraphMix(GCN)","rank_in_archive_order":6,"of":40,"metrics":{"Accuracy":"74.52 ± 0.59"},"uses_additional_data":false},{"leaderboard":"/sota/node-classification-on-citeseer-with-public-1","task":"Node Classification","dataset":"CiteSeer with Public Split: fixed 5 nodes per class","model":"GraphMix (GCN)","rank_in_archive_order":1,"of":1,"metrics":{"Accuracy":"58.55 ± 2.26"},"uses_additional_data":false},{"leaderboard":"/sota/node-classification-on-citeseer-random","task":"Node Classification","dataset":"Citeseer random partition","model":"GraphMix (GCN)","rank_in_archive_order":1,"of":1,"metrics":{"Accuracy":"76.45 ± 1.57"},"uses_additional_data":false},{"leaderboard":"/sota/node-classification-on-coauthor-cs","task":"Node Classification","dataset":"Coauthor CS","model":"GraphMix (GCN)","rank_in_archive_order":21,"of":24,"metrics":{"Accuracy":"91.83 ± 0.51"},"uses_additional_data":false},{"leaderboard":"/sota/node-classification-on-coauthor-physics","task":"Node Classification","dataset":"Coauthor Physics","model":"GraphMix (GCN)","rank_in_archive_order":13,"of":14,"metrics":{"Accuracy":"94.49 ± 0.84"},"uses_additional_data":false},{"leaderboard":"/sota/node-classification-on-cora-full-supervised","task":"Node Classification","dataset":"Cora Full-supervised","model":"GraphMix (GCN)","rank_in_archive_order":9,"of":9,"metrics":{"Accuracy":"61.8%"},"uses_additional_data":false},{"leaderboard":"/sota/node-classification-on-cora-random-partition","task":"Node Classification","dataset":"Cora random partition","model":"GraphMix (GCN)","rank_in_archive_order":1,"of":1,"metrics":{"Accuracy":"82.07 ± 1.17"},"uses_additional_data":false},{"leaderboard":"/sota/node-classification-on-cora-with-public-split","task":"Node Classification","dataset":"Cora with Public Split: fixed 20 nodes per class","model":"GraphMix","rank_in_archive_order":14,"of":36,"metrics":{"Accuracy":"83.94 ± 0.57"},"uses_additional_data":false},{"leaderboard":"/sota/node-classification-on-cora-with-public-split","task":"Node Classification","dataset":"Cora with Public Split: fixed 20 nodes per class","model":"GraphMix (GCN)","rank_in_archive_order":15,"of":36,"metrics":{"Accuracy":"83.94 ± 0.57"},"uses_additional_data":false},{"leaderboard":"/sota/node-classification-on-cora-fixed-10-node-per","task":"Node Classification","dataset":"Cora: fixed 10 node per class","model":"GraphMix (GCN)","rank_in_archive_order":2,"of":2,"metrics":{"Accuracy":"79.3"},"uses_additional_data":false},{"leaderboard":"/sota/node-classification-on-cora-fixed-5-node-per","task":"Node Classification","dataset":"Cora: fixed 5 node per class","model":"GraphMix (GCN)","rank_in_archive_order":2,"of":2,"metrics":{"Accuracy":"71.99 ± 6.46"},"uses_additional_data":false},{"leaderboard":"/sota/node-classification-on-pubmed-with-public","task":"Node Classification","dataset":"PubMed with Public Split: fixed 20 nodes per class","model":"GraphMix (GCN)","rank_in_archive_order":9,"of":37,"metrics":{"Accuracy":"80.98 ± 0.55"},"uses_additional_data":false},{"leaderboard":"/sota/node-classification-on-pubmed-with-public","task":"Node Classification","dataset":"PubMed with Public Split: fixed 20 nodes per class","model":"GCN(predicted-targets)","rank_in_archive_order":13,"of":37,"metrics":{"Accuracy":"80.42%"},"uses_additional_data":false},{"leaderboard":"/sota/node-classification-on-pubmed-random","task":"Node Classification","dataset":"Pubmed random partition","model":"GraphMix (GCN)","rank_in_archive_order":1,"of":1,"metrics":{"Accuracy":"80.72 ± 1.08"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1909.11715","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}