{"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/deep-learning-on-attributed-graphs-a-journey","title":"Deep Learning on Attributed Graphs: A Journey from Graphs to Their Embeddings and Back","arxiv_id":"1901.08296","date":"2019-01-24","proceeding":null,"authors":["Martin Simonovsky"],"abstract":"A graph is a powerful concept for representation of relations between pairs\nof entities. Data with underlying graph structure can be found across many\ndisciplines and there is a natural desire for understanding such data better.\nDeep learning (DL) has achieved significant breakthroughs in a variety of\nmachine learning tasks in recent years, especially where data is structured on\na grid, such as in text, speech, or image understanding. However, surprisingly\nlittle has been done to explore the applicability of DL on arbitrary\ngraph-structured data directly.\n  The goal of this thesis is to investigate architectures for DL on graphs and\nstudy how to transfer, adapt or generalize concepts that work well on\nsequential and image data to this domain. We concentrate on two important\nprimitives: embedding graphs or their nodes into a continuous vector space\nrepresentation (encoding) and, conversely, generating graphs from such vectors\nback (decoding). To that end, we make the following contributions.\n  First, we introduce Edge-Conditioned Convolutions (ECC), a convolution-like\noperation on graphs performed in the spatial domain where filters are\ndynamically generated based on edge attributes. The method is used to encode\ngraphs with arbitrary and varying structure.\n  Second, we propose SuperPoint Graph, an intermediate point cloud\nrepresentation with rich edge attributes encoding the contextual relationship\nbetween object parts. Based on this representation, ECC is employed to segment\nlarge-scale point clouds without major sacrifice in fine details.\n  Third, we present GraphVAE, a graph generator allowing us to decode graphs\nwith variable but upper-bounded number of nodes making use of approximate graph\nmatching for aligning the predictions of an autoencoder with its inputs. The\nmethod is applied to the task of molecule generation.","url_abs":"http://arxiv.org/abs/1901.08296v1","url_pdf":"http://arxiv.org/pdf/1901.08296v1.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":"deep-learning-on-attributed-graphs-a-journey","repo_url":"https://github.com/mys007/ecc","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"pytorch","reach":{"status":"unanswered"}}],"tasks":[{"task_slug":"graph-matching","task_name":"Graph Matching"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}