{"ID":14220,"CreatedAt":"2026-02-27T13:00:40Z","UpdatedAt":"2026-02-27T13:00:40Z","DeletedAt":null,"paper_url":"https://paperswithcode.com/paper/inductive-representation-learning-in-large","arxiv_id":"1710.09471","title":"Inductive Representation Learning in Large Attributed Graphs","abstract":"Graphs (networks) are ubiquitous and allow us to model entities (nodes) and\nthe dependencies (edges) between them. Learning a useful feature representation\nfrom graph data lies at the heart and success of many machine learning tasks\nsuch as classification, anomaly detection, link prediction, among many others.\nMany existing techniques use random walks as a basis for learning features or\nestimating the parameters of a graph model for a downstream prediction task.\nExamples include recent node embedding methods such as DeepWalk, node2vec, as\nwell as graph-based deep learning algorithms. However, the simple random walk\nused by these methods is fundamentally tied to the identity of the node. This\nhas three main disadvantages. First, these approaches are inherently\ntransductive and do not generalize to unseen nodes and other graphs. Second,\nthey are not space-efficient as a feature vector is learned for each node which\nis impractical for large graphs. Third, most of these approaches lack support\nfor attributed graphs.\n  To make these methods more generally applicable, we propose a framework for\ninductive network representation learning based on the notion of attributed\nrandom walk that is not tied to node identity and is instead based on learning\na function $\\Phi : \\mathrm{\\rm \\bf x} \\rightarrow w$ that maps a node attribute\nvector $\\mathrm{\\rm \\bf x}$ to a type $w$. This framework serves as a basis for\ngeneralizing existing methods such as DeepWalk, node2vec, and many other\nprevious methods that leverage traditional random walks.","url_abs":"http://arxiv.org/abs/1710.09471v2","url_pdf":"http://arxiv.org/pdf/1710.09471v2.pdf","authors":"[\"Nesreen K. Ahmed\", \"Ryan A. Rossi\", \"Rong Zhou\", \"John Boaz Lee\", \"Xiangnan Kong\", \"Theodore L. Willke\", \"Hoda Eldardiry\"]","published":"2017-10-25T00:00:00Z","tasks":"[\"Anomaly Detection\", \"Attribute\", \"Link Prediction\", \"Representation Learning\"]","methods":"[\"DeepWalk\", \"node2vec\"]","has_code":false}