{"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/watch-your-step-learning-node-embeddings-via","title":"Watch Your Step: Learning Node Embeddings via Graph Attention","arxiv_id":"1710.09599","date":"2017-10-26","proceeding":"NeurIPS 2018 12","authors":["Sami Abu-El-Haija","Bryan Perozzi","Rami Al-Rfou","Alex Alemi"],"abstract":"Graph embedding methods represent nodes in a continuous vector space,\npreserving information from the graph (e.g. by sampling random walks). There\nare many hyper-parameters to these methods (such as random walk length) which\nhave to be manually tuned for every graph. In this paper, we replace random\nwalk hyper-parameters with trainable parameters that we automatically learn via\nbackpropagation. In particular, we learn a novel attention model on the power\nseries of the transition matrix, which guides the random walk to optimize an\nupstream objective. Unlike previous approaches to attention models, the method\nthat we propose utilizes attention parameters exclusively on the data (e.g. on\nthe random walk), and not used by the model for inference. We experiment on\nlink prediction tasks, as we aim to produce embeddings that best-preserve the\ngraph structure, generalizing to unseen information. We improve\nstate-of-the-art on a comprehensive suite of real world datasets including\nsocial, collaboration, and biological networks. Adding attention to random\nwalks can reduce the error by 20% to 45% on datasets we attempted. Further, our\nlearned attention parameters are different for every graph, and our\nautomatically-found values agree with the optimal choice of hyper-parameter if\nwe manually tune existing methods.","url_abs":"http://arxiv.org/abs/1710.09599v2","url_pdf":"http://arxiv.org/pdf/1710.09599v2.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":"watch-your-step-learning-node-embeddings-via","repo_url":"https://github.com/benedekrozemberczki/AttentionWalk","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"watch-your-step-learning-node-embeddings-via","repo_url":"https://github.com/karthik63/attention","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"tf","reach":null}],"tasks":[{"task_slug":"graph-attention","task_name":"Graph Attention"},{"task_slug":"graph-embedding","task_name":"Graph Embedding"},{"task_slug":"link-prediction","task_name":"Link Prediction"},{"task_slug":"node-classification","task_name":"Node Classification"}],"methods":[{"method_slug":"wys","method_name":"WYS"}],"datasets_introduced":[],"methods_introduced":[{"slug":"wys","name":"WYS","full_name":"Watch Your Step"}],"results":[{"leaderboard":"/sota/node-classification-on-citeseer","task":"Node Classification","dataset":"Citeseer","model":"AttentionWalk","rank_in_archive_order":67,"of":71,"metrics":{"Accuracy":"51.5%"},"uses_additional_data":false},{"leaderboard":"/sota/node-classification-on-cora","task":"Node Classification","dataset":"Cora","model":"AttentionWalk","rank_in_archive_order":72,"of":73,"metrics":{"Accuracy":"67.9%"},"uses_additional_data":false}],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}