基于多Agent的分散式网络免疫方法研究

Propagation is one of the most ubiquitous and important dynamics of complex networks. Many phenomena such as the spread of infectious diseases, the disperse of computer viruses and the cascading of power failures can be modeled as the dynamics of different propagated events on different kinds of diffusion networks. The research of propagation and immunization aims to discover, predict and control the mechanism, the trend and the process of propagations, respectively, which is greatly important for theoretical study and demonstrates a wide range of applications. To address some fundamental issues with respect to propagation and immunization that have not been well solved so far, this project is going to implement following two tasks. (1) To propose a novel targeted immunization model being suitable for many kinds of networks with the objectives of decreasing the difference between the assumptions raised by theoretical models and real-world diffusion networks, overcoming the limitation of current targeted immunization methods that are merely competent for scale-free networks rather than uniform networks, and improving both accuracy and robustness of the prediction of propagations. (2) To propose a decentralized algorithm of targeted immunization being suitable for time-evolving, distributed and different categories of networks, which is able to obtain, under given cost constraints, an approximately optimal immunization strategy close to the global one. Not only the newly proposed algorithm can avoid relying on too much global information of networks like targeted immunization doing, but also it can obtain a critical value of immunization much better than that gotten by local immunization strategies. The implementation of this project will promote and expend the studies and the application of propagation and immunization respectively.

传播是复杂网络最普遍和最重要的动力学特性之一。传染病流行、计算机病毒蔓延和电力故障级联等诸多灾害性现象都可建模为某种流行性事件在不同类型传播网络上的传播动力学过程。网络传播和免疫机制研究旨在揭示传播规律、预测传播趋势和控制传播过程，具有重要的理论意义和广泛的应用领域。围绕传播和免疫研究还未被很好解决的若干关键问题，本项目拟开展如下研究：（1）提出适用于多种网络类型的新型目标免疫理论模型，以减小模型假定与实际传播网络结构的差异，克服现有目标免疫方法仅适用于非均匀网络而不适用于均匀网络的不足，进一步提高传播过程预测的准确性和可靠性；（2）提出适用于动态、分布和多种类型网络的分散式目标免疫算法，能够基于局部网络信息在免疫成本约束下获得接近全局最优的免疫效果，既能避免目标免疫使用过多全局网络信息的不足，又能得到优于局部免疫策略的临界值。本项目的实施对促进复杂网络传播和免疫的研究与应用具有重要意义。

传播是复杂网络最普遍和重要的动力学特性之一。传染病流行、计算机病毒蔓延和电力故障级联等诸多灾害性现象都可建模为某种流行性事件在不同类型传播网络上的传播动力学过程。网络传播和免疫机制研究旨在揭示传播规律、预测传播趋势和控制传播过程，具有重要的理论意义和广泛的应用领域。围绕传播和免疫研究还未被很好解决的若干关键问题，本项目从隐含传播网络推断和流行病主动监控等几个方面开展研究，提出了基于早期检测和主动监控的网络免疫研究新思路和相关算法。在该项目的资助下，项目组累计发表标注资助论文18篇，其中被SCI收录5篇，CCF-A类文章6篇。