面向大数据的统计分布式计算及隐私保护的理论与方法

本项目聚焦于分布式计算的前沿探索，是一项解决大数据管理决策问题的基础性研究。首先对于非光滑损失函数以及U统计量类型的损失函数设计非迭代且高效通讯的分布式算法，使得提出的方法可以有效地应用于大规模数据集场景的分析与建模。同时给出算法中参数估计的统计性质，以使方法在实际应用中具有良好的解释性，并能进行有效的统计推断。最后基于大数据分布式算法，围绕大数据分析中备受关注的数据安全问题开展讨论，提出科学合理的隐私保护技术与方法，最终达到使以上研究的分布式计算方法既能满足数据安全要求，又不破坏原有的数据通讯以及统计推断效率的目的，并将其应用到管理、金融、生物医学、高精密行业的实际场景，为管理决策提供咨询建议和指导。

In the big data era, the research focuses on the frontiers of distributed computing that solves the basic problems in management and decision-making, which are effectively applicable in the analysis and modeling of large-scale data set scenarios. More specifically, we develop the communication-efficient and non-iterative distributed algorithm for non-smooth loss function and complex loss function of U-type statistics, respectively. At the same time, the properties of parameter estimation under communication-efficient distributed computing are given, so that the method has good interpretability in the applications and can carry out effective statistical inference. Finally, we discuss the data security issues that are of great concern in big data analysis based on the distributed algorithm and propose the scientific and reasonable data privacy protection technologies to achieve the balance between the requirements of data privacy and the efficiency of the computing and inference. The above methods will be widely applied in the areas of management, finance, biomedicine, and high-precision industries to provide guidance for management.