DeepWalk: Online Learning of Social Representations
Bryan Perozzi Rami Al-Rfou Steven Skiena Stony Brook University Department of Computer Science Stony Brook University Department of Computer Science Stony Brook University Department of Computer Science ralrfou
Abstract
We present DeepWalk, a novel approach for learning latent representations of vertices in a network. These latent representations encode social relations in a continuous vector space, which is easily exploited by statistical models. DeepWalk generalizes recent advancements in language modeling and unsupervised feature learning (or deep learning) from sequences of words to graphs.
中文速览
在社交网络中,每个节点(用户)究竟跟谁"亲近"、属于哪个社群,传统方法要么依赖全局网络结构、要么把标签信息混进特征里,标注数据一旦稀缺就表现很差。DeepWalk 把这个问题转化成一个"语言建模"任务:对每个节点做短距离随机游走(random walk),把游走序列当成句子,再借鉴 Word2Vec 的训练方式,让网络中结构相近的节点在低维连续向量空间里也彼此靠近,从而学出每个节点的隐式表示(latent representation)。在 BlogCatalog、Flickr、YouTube 等多个真实社交网络的多标签分类实验中,DeepWalk 比那些能看到全局网络信息的基线方法高出最多 10% 的 F₁ 分数,甚至在训练数据减少 60% 的情况下仍能压过所有对比方法。这项工作的重要性在于,它首次将深度学习中的无监督表示学习引入网络分析,所学特征与标签无关、可复用于多个任务,算法本身支持在线更新和并行化,为后续图表示学习研究奠定了基础。
原文 arXiv:1403.6652;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1403.6652v2