抽样对复杂网络生长机制的影响

运用基于机器学习理论开发的网络机制辨识方法，对不同抽样方案得到的抽样网络生长机制进行识别，将结果与完全网络比较,以分析抽样对网络机制的影响.研究表明，不同抽样方案对网络生长机制具有不可忽视的影响.在抽样绩效方面，中枢链式抽样具有较好的网络机制保持能力；而对比网络结构特征与网络机制，中枢抽样和链式抽样对网络机制的影响更显著.此外，各抽样方案对网络机制的影响随网络类型的不同有一定差异，这表明在实践中有必要根据网络类型选择恰当的抽样方法.

Effect of Sampling on Growth Mechanisms of Complex Networks

The problem of whether the topological structure of network can be kept well when data is collected incompletely is concerned. The current focus on the effect of sampling on macro statistical properties has been extended to consideration of its effect on micro network mechanisms. A mechanism-inferring method of networks exploited from machine learning theory was used to identify the mechanisms of sampled networks obtained by different sampling methods. The results were compared with the original networks to analyze the impact of sampling on network mechanisms. It is found that different sampling methods have a nontrivial influence on growth mechanisms of networks. In conclusion, the hub-chain sampling strategy can keep mechanisms better than others. Meanwhile, hub sampling and chain sampling have more remarkable influences on network mechanisms than their effects on structural properties. In addition, there are some differences in the effects of sampling for different types of networks, which means in practice we should select suitable sampling method according to the research object of network to improve the accuracy of network building more effectively.