III: Small: Analysis and Models of Social Network Structure, Growth and Dynamics

在线社交网络（OSN），如Facebook和LinkedIn，是大量互联网用户之间进行通信和交互的有价值的基础设施。多年来，研究人员一直试图回答有关这些复杂网络的形成，它们的持续演变，内部结构的形成以及不同时间尺度的变化的基本问题。由于回答这些问题需要大规模的真实的动力学数据集，因此大多数先前的研究都受到缺乏数据的严重限制。OSN提供商已授权主要研究者访问一个社交网络超过2年的独特详细和完整的动态跟踪。目标是挖掘和分析网络动力学的痕迹，以验证现有模型并指导细粒度网络动力学的新模型。目标包括在网络增长的不同阶段的优先连接模型的分析，在时间和图拓扑结构的细粒度开发新的网络动力学模型，并探索由新的图动力学度量驱动的应用程序，这项工作有可能极大地改变我们对在线社交网络中的动态的理解。通过采用经验的、数据驱动的方法进行网络建模，他们可以揭示传统的网络动态模型是如何偏离真实情况的。此外，他们正在开发经验模型，这些模型在小规模下更有效地准确预测网络事件。赵和郑都在为女性和少数民族学生的教育和推广项目上投入了大量资金：在他们的实验室里，女性学生和博士后的数量往往超过男性同行。PI将把他们的研究结果传播给他们在Renren和LinkedIn的合作者，并通过现有的技术联系和非正式访问/会谈与Twitter、Zynga、Facebook和Google的研究人员分享研究结果。欲了解更多信息，请访问项目网页（http：//sandlab.cs.uchicago.edu/dynamics/）。

Online social networks (OSNs) such as Facebook and LinkedIn are valuable infrastructures for communication and interactions between a large volume of Internet users. For years, researchers have been trying to answer fundamental questions about the formation of these complex networks, their ongoing evolution, formation of internal structures, and change at different time scales. Since answering these questions requires real dynamics datasets at scale, most prior studies have been significantly constrained by a lack of data. The Principal Investigators have been granted access by an OSN provider to a uniquely detailed and complete trace of dynamics over 2+ years of a social network. The goal is to mine and analyze the traces of network dynamics to validate existing models and guide new models for fine grain network dynamics. Objectives include analysis of the preferential attachment model at different stages of network growth, developing new models of network dynamics at fine granularity in both time and graph topology, and explorations of applications driven by novel metrics of graph dynamics.The work has the potential to dramatically change our understanding of dynamics in online social networks. By taking an empirical, data-driven approach to network modeling, they can shed light on how traditional models of network dynamics deviate from ground truth. In addition, they are developing empirical models that are more effective at accurately predicting network events at small scales. Both PIs Zhao and Zheng are heavily invested in educational and outreach programs for female and minority students: female students and postdocs often outnumber male counterparts in their lab. The PIs will disseminate their results to their collaborators atRenren and LinkedIn, and also share results with researchers at Twitter, Zynga, Facebook and Google through existing technical contacts and informal visits/talks. For further information, please see the project webpage (http://sandlab.cs.uchicago.edu/dynamics/).