{"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/every-node-counts-self-ensembling-graph","title":"Every Node Counts: Self-Ensembling Graph Convolutional Networks for Semi-Supervised Learning","arxiv_id":"1809.09925","date":"2018-09-26","proceeding":null,"authors":["Yawei Luo","Tao Guan","Junqing Yu","Ping Liu","Yi Yang"],"abstract":"Graph convolutional network (GCN) provides a powerful means for graph-based\nsemi-supervised tasks. However, as a localized first-order approximation of\nspectral graph convolution, the classic GCN can not take full advantage of\nunlabeled data, especially when the unlabeled node is far from labeled ones. To\ncapitalize on the information from unlabeled nodes to boost the training for\nGCN, we propose a novel framework named Self-Ensembling GCN (SEGCN), which\nmarries GCN with Mean Teacher - another powerful model in semi-supervised\nlearning. SEGCN contains a student model and a teacher model. As a student, it\nnot only learns to correctly classify the labeled nodes, but also tries to be\nconsistent with the teacher on unlabeled nodes in more challenging situations,\nsuch as a high dropout rate and graph collapse. As a teacher, it averages the\nstudent model weights and generates more accurate predictions to lead the\nstudent. In such a mutual-promoting process, both labeled and unlabeled samples\ncan be fully utilized for backpropagating effective gradients to train GCN. In\nthree article classification tasks, i.e. Citeseer, Cora and Pubmed, we validate\nthat the proposed method matches the state of the arts in the classification\naccuracy.","url_abs":"http://arxiv.org/abs/1809.09925v1","url_pdf":"http://arxiv.org/pdf/1809.09925v1.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":"every-node-counts-self-ensembling-graph","repo_url":"https://github.com/RoyalVane/SEGCN","is_official":1,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":"classification","task_name":"General Classification"},{"task_slug":"node-classification","task_name":"Node Classification"}],"methods":[{"method_slug":"dropout","method_name":"Dropout"},{"method_slug":"gcn","method_name":"GCN"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/node-classification-on-citeseer-with-public","task":"Node Classification","dataset":"CiteSeer with Public Split: fixed 20 nodes per class","model":"SEGCN","rank_in_archive_order":15,"of":40,"metrics":{"Accuracy":"73.4 ± 0.7"},"uses_additional_data":false},{"leaderboard":"/sota/node-classification-on-cora","task":"Node Classification","dataset":"Cora","model":"SEGCN","rank_in_archive_order":39,"of":73,"metrics":{"Accuracy":"83.5% ± 0.4%"},"uses_additional_data":false},{"leaderboard":"/sota/node-classification-on-cora-fixed-20-node-per","task":"Node Classification","dataset":"Cora: fixed 20 node per class","model":"SEGCN","rank_in_archive_order":4,"of":9,"metrics":{"Accuracy":"83.5 ± 0.4"},"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":"SEGCN","rank_in_archive_order":24,"of":37,"metrics":{"Accuracy":"78.9 ± 0.7"},"uses_additional_data":false}],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}