复杂网络与机器学习融合的研究进展  

A REVIEW OF COMBINING COMPLEX NETWORKS AND MACHINE LEARNING

作　　者：李泽荃 杨曌 刘嵘 李靖 Li Zequan;Yang Zhao;Liu Rong;Li Jing(School of Management, North China Institute of Science and Technology, Beijing 101601, China;School of Safety Engineering, North China Institute of Science and Technology, Beijing 101601, China;School of Computer, North China Institute of Science and Technology, Beijing 101601, China)

机构地区：[1]华北科技学院管理学院,北京101601 [2]华北科技学院安全工程学院,北京101601 [3]华北科技学院计算机学院,北京101601

出　　处：《计算机应用与软件》2019年第4期10-28,62,共20页Computer Applications and Software

基　　金：中央高校基本科研业务费资助项目(3142018050;3142017105;3142017088)

摘　　要：近年来,随着大数据技术的进步,复杂网络与机器学习的交叉研究越来越受到众多学者的关注。复杂网络是自然界中众多复杂系统的抽象描述,主要以统计物理的角度研究系统的演化;机器学习又称为统计学习方法,主要研究从大量数据样本提取特征并建立模型。简要综述复杂网络领域主要的网络演化模型、常用统计度量方法以及网络上的动力学过程和机器学习领域内三种基本的学习技术;从交叉应用的两个角度,即基于复杂网络的机器学习方法和基于机器学习的复杂网络信息挖掘,详细对比了各种方法的计算思路。在此基础上,提出目前学界重点关注的两类问题,并展望了若干开放性挑战。In recent years, with the progress of big data technology, more and more scholars pay attention to the interdisciplinary research on complex networks and machine learning. Complex network is an abstract description about many complex systems in nature, which studies the evolution of the system mainly in the perspective of statistical physics. Machine learning is also called statistical learning method , and its main research is to extract features from a large number of data samples and to establish a model. This paper briefly reviewed the main networks evolution models in the complex network, the commonly-used statistical measurements, dynamics on the networks and three basic learning techniques in the field of machine learning. The calculation ideas of various methods were compared in details from two perspectives of cross-application, i. e. machine learning based on complex network and information mining based on machine learning. On this basis, two important issues in the academic circle were put forward and some open problems are looked forward to.

关 键 词：复杂网络 机器学习 社团检测 链路预测 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]