【作者】 彭盾；

【导师】 曾德明；

【作者基本信息】 湖南大学 ， 企业管理， 2010， 博士

【摘要】 高技术企业创新的复杂性决定了企业需不断地向外扩展自身R&D边界以增加自身知识存量,在价值最大化、交易成本最小化的指导下,企业便会采用最合理的治理结构规制、保护合作交易。从“关系”的视角去看待企业的创新系统,能更好地理解单个主体之间的合作行为,进而能推导出整个网络的结构与功能。网络自生成以来,在不同机制作用下,向着不同的结构、不同的内涵演化。本文构建了演化模型,考虑不同的网络初始结构,不同的演化机制下网络最终的拓扑结构以及功能。近年来,复杂性理论的最新研究成果,尤其是“复杂网络”理论与方法的发展,使得运用复杂性理论来研究创新网络成为可能。在复杂网络理论研究不断深入的同时,对复杂网络理论的应用研究已开始从计算机控制学科快速扩散到社会经济管理领域。为寻找到网络演化的初始拓扑结构,本文借助静态均衡分析框架,利用有效性和稳定性两个原则构建模型,得到网络的初始拓扑结构是“星型网络”和“全连通网络”。通过对大量产业集群的分析,这两种网络普遍存在于现实经济社会之中。现实中,还存在一种复杂关系形式的创新网络,这些发现都将为文章的演化模型构建奠定坚实基础。关于创新网络的演化研究,必然还要归纳演化机制。高技术产业的特性—主体间知识资源高度互补性、网络发展不确定性以及社会资本全局重要性,再加上技术标准等权变因素要求高技术企业需不断通过跨组织学习获得外部知识资源,从而使得网络规模不断扩大,网络呈现增长机制；企业间的跨组织学习在地域邻近性、关系邻近性以及认知邻近性的影响下,企业在选择合作伙伴时会出现“择优连接机制”,也即企业会选择知识资源多的企业,或者有良好合作历史、声誉高的企业。在此基础上,本文分别构建了星形网络以及全连通网络的加权网络演化模型。当初始网络为星形网络时,在网络发展的过程中,非中心点之间同样也会发生合作创新行为,那么整个网络将演化成无标度的完全图,或者这个网络中少量的节点会含有大量的边,我们把这些节点称为“富有节点”,他们倾向于彼此相互连接,构成“富人俱乐部”。当网络的初始结构是全连通网络时,在边权机制驱动下,网络最终会演变成无标度网络。创新主体除通过合作从外获得知识资源外,自身也会通过R&D获得知识以增强自身在网络中的影响力,主体呈现自适应学习机制。尽管初始结构不同,但网络最终都会演变成无标度网络,本文建立一个统一的演化改进模型。结果显示,从知识流动的视角来看待技术创新网络拓扑结构的演化行为,可以分为区域集群和远程连接,分别对应于节点企业技术创新合作中的基于创新机会寻找。此外,网络拓扑结构最终会演变成小世界网络,同时还具有无标度性。研究技术创新网络结构最终的目的是发现网络拓扑结构对网络上知识流动效率的影响。在同步思想指导下,考虑到知识扩散过程中的知识再次创新,本文构建了知识扩散的网络同步模型。此模型要求网络结构呈现多中心结构,然这种结构对于网络“蓄意攻击”呈现脆弱性。在考虑到同步以及鲁棒性情况下,构建了同步优先技术创新网络。这种网络即时在去除很多中心的情况下,网络依然能够保持连通性,具有很强鲁棒性。根据对中国基因工程制药企业创新网络的实证发现,2004-2008年间,创新网络具有无标度性。企业技术创新网络的主体在增加,但是平均最短路径却变小了。这说明,合作关系在进一步加强。这些结论对于我国基因制药工程产业的启示,一个比较不错的战略就是形成全国性的小世界网络,不但具有较短平均最短路径还具有较高的集群系数。这一方面,有利于知识扩散,另一方面又有利于知识创新。更多还原

【Abstract】 Complexity of innovation determines High-tech firms expanding their R&D boundaries for acquisition of external knowledge resource, thus to accumulate their knowledge stock. In order to maximum value creation and minimum transaction cost, firms would select a rational governance structure to protect this cooperative transaction. To understand the potential complementarities between individual transactions in a firm’s portfolio (or a network) we further consider another theoretical perspective on firm strategy:the relational view of the firm, which emphasizes the whole portfolio of transactions as a unique resource of the firm. This paper constructs several evolutional models, for considering many original topology structures and different evolutional mechanism, to analyze the ultimate network topology structure and function.Recent years, there has some important broke through in complex theory area, especially the development of "complex networks". As the development of complex networks theory, at meanwhile, researchers expand its application areas, for example, this methodology can be used as a useful tool for studying innovation networks.In order to build network evolutional model, we must find the original topology structure. Therefore, this paper constructs a static equilibrium analytical framework according to two principles:effective principle and stable principle. As a result, "star network" and "homogeneous network" are found as the original network structure. In the real world, these two network structure exist wildly. Out of our mind, there exists a complex innovation network in some industry clusters. But these findings are all useful for constructing our evolutional models.In order to build network evolutional model, we should also conclude evolutional mechanisms. The characteristics of high-tech industry:complementary of individual firm’s knowledge resource, uncertainty of network development and importance of social capital for whole network, and many other contingency factors ask for the high-tech firms to outsourcing external knowledge resource through inter-organizational learning. This makes network expand its scale, thus network appears growth mechanism; in another way, there proximities (cognitive proximity, relational proximity and geographical proximity) affect this learning process. This makes network present "preferential attachment" mechanism when firms choose their cooperative partners. In others words, firms are ease to choose these who possess higher knowledge stock, or those who have good cooperative history, or those who has good cooperative reputation as their partner.Base on this, this paper built two evolutional models separately. In this first situation, the original network structure is "star network". In the network development process, there are cooperative transactions between un-core nodes; thus the network will be a free-scale network, or few nodes possess huge verticals, we called these nodes "rich-node"; these nodes are also ease to connect with each other, they form "rich-club". In the second model, the original network structure is "homogeneous network". This network will be a free-scale network at finally.Individual firm also can gain knowledge by their self through R&D besides external knowledge resource acquisition, this also can raise his influential capacity in the innovation networks, and thus individual firm presents adaptive learning mechanism in the evolution process. Although the original network structures are different, the final outcome is the same; so here, we built a unified improved evolutional model. The outcome shows that we can conclude evolutional behaviors from knowledge flow perspective, network nodes using clustering and distant linkage to find innovative opportunities. Beside this, the network will be a small world network, also have free-scale character.The purpose of studying network topology structure is to find the impact of network structure on the knowledge diffusion on the innovation networks. According to synchronization theory, considering knowledge creation in the process of knowledge diffusion, this paper builds a knowledge diffusion network synchronization model. This model requires the networks have multi core nodes. But, this structure is fragile to "deliberate attack". For comprehensive considering of robustness and synchronization, this paper builds a synchronous preference innovation networks; This network possesses higher robustness though some core nodes deviate from this network.This paper also did empirical study on China gene engineering pharmaceutical firms, in the period 2004-2008, this network is a free-scale network. During this period, many new nodes enter into this network, but the average path length is becoming shorter. These findings can be used to give suggestions for China gene engineering pharmaceutical industry; they should build a national small world network. This network will have shorter average path length and high cluster coefficient; these are good for knowledge diffusion and knowledge creation.更多还原