协作网络演化动力学模型研究

发布时间：2018-04-23 08:02

  本文选题：协作网络 + 动力学模型　； 参考：《大连理工大学》2016年博士论文

【摘要】：现实中,由个体间通过自发调整协作关系而形成的协作网络(如科研协作网络、开源软件开发者网络等)在求解复杂社会经济问题方面具有个体难以比拟的优势。大量实证研究表明这类协作网络求解复杂问题的能力同其成员问不断协调的过程,以及所引发宏观网络的结构与演化过程密切相关,并且往往具有显著的社区结构与小世界特征。然而,虽然在过去的二十余年学界针对协作网络的社区与小世界结构的形成及演化开展了大量研究,但迄今人们对于协作网络的多社区小世界结构的形成机理及演化规律的认识还有待进一步深化。特别是,对于多社区小世界网络的形成,目前学界主要归因于社会资本视角下的结构镶嵌与结构洞两种协作关系形成机制的组合；而另有实证研究表明协作关系的建立与个体基于属性状态相似性(或差异性)的倾向性交互密切相关。本文认为个体基于属性状态相似性(或差异性)的倾向性交互对于协作关系的形成,继而促使协作网中络多社区小世界状态涌现的作用不应忽视,尤其是Homophily、Heterophily、以及社会学习机制共同作用下协作网络的结构及演化问题还值得进一步探讨。因此,在对已有相关研究进行梳理的基础上,本文采用动力学模型研究与数据分析相结合的方式,对上述三种机制下协作网络的结构及演化进行了考察。首先,本文检测了结构镶嵌与结构洞机制对于网络结构形成的作用,并得出这两种机制下模型生成网络难以形成稳定的小世界状态,且不具有显著的社区结构。进而,本文构建了基于Homophily、Heterophily、以及社会学习机制的协作网络演化动力学模型,在网络规模恒定的情景下,对这三种机制与网络结构形成之间的关系进行了全面的考察。数值实验结果显示,Homophily机制促使相似个体汇聚,继而形成社区结构；Heterophily机制引发跨社区连接的建立；社会学习促使相似个体属性状态趋同,导致由相似个体构成的社区内部边密度增大,对社区结构具有强化作用。因此,在这三种机制的共同作用下,特别是Homophily与Heterophily机制的一定组合下,模型生成网络呈现出稳定的多社区小世界状态,这一结果在填补了现有研究对于这些机制缺乏综合性探讨的同时,从个体基于属性状态相似性(或差异性)的倾向性交互视角阐明了协作网络中多社区小世界结构的形成机理。其次,在明确了Homophily、Heterophily、以及社会学习机制对于网络结构形成不同作用的基础上,本文进一步构建了基于Homophily与Heterophily组合机制的协作网络演化动力学模型,研究了在网络规模不断增长的情景下模型生成网络的结构及演化过程。研究结果表明,在模型生成网络由分散状态逐渐汇聚为“中央—边缘”结构的演化过程中,可根据“中央”(最大连通子图)的结构变化将模型生成网络的结构依次区分为分散网络、链式结构、以及多社区小世界网络三种状态。这一结果不仅弥补了相关工作中对于协作网络演化规律缺少明确理论研究的不足,也对协作网络结构及演化相关理论的完善具有推动作用。最后,本文将“合作演化”与“复杂网络”研究领域的合著网络、以及Linux内核开发参与者网络作为现实协作网络数据分析实例,对其结构及演化过程进行了考察。分析结果显示所选取的三个现实网络均存在由分散到汇聚为“中央—边缘”结构的演化过程,并且各自的最大连通子图均会由一个小规模的聚簇逐渐形成一种链式结构,最终演化为兼具多社区小世界结构与层次结构的网络,相应三个现实网络在演化过程中依次呈现出分散网络、链式结构、以及多社区小世界网络三种不同的状态。这一结果不仅说明本文所考察的这类协作网络可能具有共性的结构特征与演化模式,同时也是对已有数据分析工作的很好补充。进而,数据分析结果与动力学模型研究具有较好的一致性,进一步说明本文所提出的协作网络演化动力学模型可以反映现实协作网的结构及演化过程,并且个体基于属性状态相似性(或差异性)的倾向性交互可能是导致现实协作网络呈现多社区小世界结构的重要动因,从而为相关管理实践活动提供了理论支持。
[Abstract]:In reality, a collaborative network (such as scientific research collaboration network, open source software developer network, etc.) formed by the spontaneous adjustment of cooperative relations between individuals (such as scientific research collaboration network, open source software developer network) has an unparalleled advantage in solving complex social and economic problems. A large number of empirical studies show that the ability of this kind of collaborative network to solve complex problems is consistent with its members. The process, as well as the structure and evolution process of the resulting macro network, is closely related to the evolution of the community, and often has significant community structure and small world characteristics. However, although a lot of research has been carried out on the formation and evolution of the community and the small world structure in the past twenty years, there has been a lot of collaborative networks to date. The understanding of the formation mechanism and evolution of the community small world structure remains to be further deepened. Especially, for the formation of the multi community small world network, the current academic circles are mainly attributable to the combination of two cooperative relationships between structural inlay and structural cavities under the perspective of social capital; and another empirical study shows the construction of the cooperative relationship. The tendency of sexual intercourse based on attribute state similarity (or difference) is closely related to each other. This paper argues that the tendency of individual based on attribute state similarity (or difference) to form mutual cooperation is to promote the role of the small world state of the complex community in the cooperative network, especially Homophily, Heterophily, The structure and evolution of the cooperative network under the joint action of the social learning mechanism is still worth further discussion. Therefore, on the basis of combing the existing related research, the structure and evolution of the cooperative network under the three mechanisms are investigated by the combination of dynamic model and data analysis. In this paper, the effect of structural inlay and structure hole mechanism on the formation of network structure is detected, and it is concluded that the model generation network is difficult to form a stable small world state under these two mechanisms, and does not have a significant community structure. Furthermore, this paper constructs a collaborative network evolution dynamics based on Homophily, Heterophily, and social learning mechanism. The study model, under the constant network scale, investigates the relationship between the three mechanisms and the formation of the network structure. The results of numerical experiments show that the Homophily mechanism promotes the convergence of similar individuals and then forms a community structure; the Heterophily mechanism leads to the establishment of cross community connections; social learning encourages similar individuals. Therefore, under the joint action of the three mechanisms, especially the combination of the Homophily and the Heterophily mechanism, the model generation network presents a stable multi community small world state under the joint action of these three mechanisms, and this result is filling the present situation. There is a lack of comprehensive discussion on these mechanisms. At the same time, the formation mechanism of the multi community small world structure in the cooperative network is clarified from the individual similarity (or difference) based on the attribute state similarity (or difference). Secondly, it is clear that Homophily, Heterophily, and social learning mechanism have different effects on the network structure. On the basis of this, this paper further constructs a cooperative network evolutionary dynamics model based on the combination of Homophily and Heterophily, and studies the structure and evolution process of the model generation network under the growing network scale. The results show that the model generation network is gradually converged to "central edge" in the model generation network. In the process of structure evolution, the structure of the model generation network can be divided into three states, namely, decentralized network, chain structure, and multi community small world network according to the structure change of "central" (the most Dalian Tong Zi map). This result not only makes up for the lack of clear theoretical research on the evolution law of cooperative network in the related work. It also promotes the improvement of the cooperative network structure and evolution related theories. Finally, this paper examines the co authoring network of "cooperative evolution" and "complex network" and the Linux kernel development participant network as a practical collaborative network data analysis example, and investigates its structure and evolution process. The three real networks selected are evolved from decentralization to convergence to the "central edge" structure, and each of the most Dalian maps will gradually form a chain structure by a small cluster of clusters, which eventually evolved into a network with a multi community small world structure and a hierarchical structure, and the corresponding three real networks. In the process of evolution, there are three different states of distributed network, chain structure, and multi community small world network. This result not only shows that this kind of cooperative network may have common structural features and evolution patterns, but also a good supplement to the existing data analysis work. The study of fruit and dynamics model has good consistency. It further illustrates that the proposed cooperative network evolutionary dynamics model can reflect the structure and evolution process of the real collaboration network, and the tendency of the individual based on the similarity of attribute state (or difference) may cause the real collaboration network to present a small community of small communities. The important motivation of structure provides theoretical support for related management practices.

【学位授予单位】：大连理工大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：O157.5
，

本文编号：1791116