【摘要】 大数据是一笔越来越重要并不断快速增长的财富,合理利用这一财富的关键是有效的分析手段.大数据中一大类数据是由复杂网络代表的实际动力学系统产生的,其中网络各个单元的输出数据可以测量,但产生数据的网络结构却不为所知;而了解这些网络结构对我们理解、预测和控制实际系统功能极为重要.因此,从分析网络数据出发揭示网络结构的重构问题就成为数学物理特别是统计物理以及一系列交叉领域对网络研究的核心问题之一.网络重构的重要性还来源于解决实际网络重构中所面对的各种困难的理论要求.网络结构的复杂性、网络节点动力学的非线性、未知噪声对网络动力学演化数据的影响以及测量中有效数据的缺失等都是在实际网络重构中要面对的常见且非常重要的困难.本综述介绍并讨论了如何有效克服这些困难的方法,特别是通过数据扩张充分利用数据信息的方法,针对不同的系统特征和重构任务选择合适的关联量计算方法,以及利用噪声帮助克服重构困难的方法等.网络重构研究将逐步解决实际复杂系统重构问题并引起复杂网络相关学者越来越多的关注和研究兴趣.更多还原

【Abstract】 In recent decades, large scale of data sets have been accumulated in various and wide fields, in particular in social and biological systems. There are massive amounts of data available for utilization; and for really using these data it is crucial to find effective methods of analyses. Many practical systems can be described by dynamic networks, for which modern technique can measure their outputs, and accumulate extremely rich data. Nevertheless, the network structures producing these data are often deeply hidden. The structure and intensity of couplings between the network elements(say, network nodes) play key roles in determining the features of network dynamics. Therefore, the problem of network reconstruction,i.e., exploring unknown network structures by analyzing available output network data, has attracted significant interest in many interdisciplinary fields in recent decades. The significance of network reconstruction is due to not only the extreme importance of black-box manifestation in practice, but also serious difficulties, which are theoretically challenging, and appearing in various aspects of real-world systems, such as complexity of network structures, strong nonlinearity of network dynamics, diverse and unknown impacts from the interiors of nodes and the externals of networks, i.e., presence of noises,many hidden and not measurable nodes in networks. In this review, we introduce some effective methods dealing with the above difficulties, such as exploring information of available data by searching rich and diverse data features; seeking different types of correlations aiming at treating different difficulties; finding intelligent ways to remove negative and utilize active effects of noises, and so on. Our approach offers some possibilities towards understanding of complex networked systems in more complicated and difficult conditions and has important implications for the reconstruction tasks of many practical networks.更多还原