{"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/ddgk-learning-graph-representations-for-deep","title":"DDGK: Learning Graph Representations for Deep Divergence Graph Kernels","arxiv_id":"1904.09671","date":"2019-04-21","proceeding":null,"authors":["Rami Al-Rfou","Dustin Zelle","Bryan Perozzi"],"abstract":"Can neural networks learn to compare graphs without feature engineering? In\nthis paper, we show that it is possible to learn representations for graph\nsimilarity with neither domain knowledge nor supervision (i.e.\\ feature\nengineering or labeled graphs). We propose Deep Divergence Graph Kernels, an\nunsupervised method for learning representations over graphs that encodes a\nrelaxed notion of graph isomorphism. Our method consists of three parts. First,\nwe learn an encoder for each anchor graph to capture its structure. Second, for\neach pair of graphs, we train a cross-graph attention network which uses the\nnode representations of an anchor graph to reconstruct another graph. This\napproach, which we call isomorphism attention, captures how well the\nrepresentations of one graph can encode another. We use the attention-augmented\nencoder's predictions to define a divergence score for each pair of graphs.\nFinally, we construct an embedding space for all graphs using these pair-wise\ndivergence scores.\n  Unlike previous work, much of which relies on 1) supervision, 2) domain\nspecific knowledge (e.g. a reliance on Weisfeiler-Lehman kernels), and 3) known\nnode alignment, our unsupervised method jointly learns node representations,\ngraph representations, and an attention-based alignment between graphs.\n  Our experimental results show that Deep Divergence Graph Kernels can learn an\nunsupervised alignment between graphs, and that the learned representations\nachieve competitive results when used as features on a number of challenging\ngraph classification tasks. Furthermore, we illustrate how the learned\nattention allows insight into the the alignment of sub-structures across\ngraphs.","url_abs":"http://arxiv.org/abs/1904.09671v1","url_pdf":"http://arxiv.org/pdf/1904.09671v1.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":"ddgk-learning-graph-representations-for-deep","repo_url":"https://github.com/google-research/google-research/tree/master/graph_embedding/ddgk","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":0,"framework":"tf","reach":null}],"tasks":[{"task_slug":"feature-engineering","task_name":"Feature Engineering"},{"task_slug":"graph-attention","task_name":"Graph Attention"},{"task_slug":"graph-classification","task_name":"Graph Classification"},{"task_slug":"graph-similarity","task_name":"Graph Similarity"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/graph-classification-on-dd","task":"Graph Classification","dataset":"D&D","model":"DDGK","rank_in_archive_order":4,"of":53,"metrics":{"Accuracy":"83.14%"},"uses_additional_data":false},{"leaderboard":"/sota/graph-classification-on-mutag","task":"Graph Classification","dataset":"MUTAG","model":"DDGK","rank_in_archive_order":13,"of":74,"metrics":{"Accuracy":"91.58%"},"uses_additional_data":false},{"leaderboard":"/sota/graph-classification-on-nci1","task":"Graph Classification","dataset":"NCI1","model":"DDGK","rank_in_archive_order":64,"of":69,"metrics":{"Accuracy":"68.1%"},"uses_additional_data":false},{"leaderboard":"/sota/graph-classification-on-ptc","task":"Graph Classification","dataset":"PTC","model":"DDGK","rank_in_archive_order":26,"of":37,"metrics":{"Accuracy":"63.14%"},"uses_additional_data":false}],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}