网络表示学习算法的研究现状与进展

Research Status and Progress of Network Representation Learning Algorithms

网络数据可以用来描述事物及事物之间的复杂联系。真实世界中诸如社交、文献引文等大型网络,结构复杂、规模大,节点通常包含丰富的内容信息,随之带来的网络高维、稀疏等特点给机器学习带来前所未有的挑战。网络表示学习融合网络结构信息和节点内容信息将网络节点映射到低维向量空间,能够提高机器学习的效率。根据网络特点对不同类型的网络表示学习进行了介绍与总结。

Network data can be used to describe things and complex connections between things.In the real world,large networks such as social networking and literature citations have complex structures and large scales,and nodes usually contain rich content information.The high-dimensional and sparse characteristics of the network have brought unprecedented challenges to machine learning.Network representation learning integrates network structure information and node content information to map network nodes to low-dimensional vector spaces,which can improve the efficiency of machine learning.This article introduces and summarizes different types of network representation learning based on network characteristics.