基于命名数据网络的分布式推理研究

发布时间：2018-03-01 15:26

本文关键词： 动态分布数据知识发现分布式推理命名数据网络语义整合　出处：《湖南科技大学》2014年硕士论文　论文类型：学位论文

【摘要】：随着网络和通信技术的发展，互联网络已经演变成为如今面向内容分发与服务提供的普适信息基础设施。而在大规模分布和动态的数据源中智能地发现知识和规则等是一个极具挑战的问题。面向数据（data-oriented）、以内容或信息为中心的网络协议体系实现了从“机器互联”到“信息互联”的转变。本研究面向命名数据网络（NamedDataNetworking）体系结构，针对庞大的位置无关数据命名空间，，主要研究分布、自组织的语义推理方法、算法模型和系统实现方法。其中包括：第一，研究适用命名数据网络环境的、基于用户驱动的语义推理机制。命名数据网络是将内容的名字与其位置相分离，以实现动态分布的网络环境下基于内容名字的路由。（1）在动态分布的网络环境下，为了发现所有与用户需求相关的、潜在的数据源，研究用户推理请求的语义转发过程，在转发节点中设计具有推理能力的转发引擎；（2）针对分布的推理结果的返回和汇聚过程，研究推理结果的语义整合和集聚机制及实现方法；（3）由于整个推理过程需要在分布、自组织的环境下进行，因此本文还研究提出具有动态自适应性和高效的推理机制和算法。第二，研究基于请求驱动和路由过程中的推理规则。在动态分布的网络环境下进行语义推理需要简单、有效。首先，针对分布环境下用户的推理需求，提出基于请求驱动的推理规则表示方法；然后，基于层次的内容命名机制和路由索引的特征，提出请求的转发推理机制，以发现所有潜在的知识源并将请求转发给它们；最后，在汇聚点基于聚集的返回结果，研究汇聚结果的语义整合规则，以达到精练、整合推理结果的目标，并研究推理请求的演化规则，以根据已获取的部分知识更新请求，从而发现用户需要的更多潜在知识。第三，研究基于命名数据网络的、动态分布的语义转发和整合推理的系统实现方法。在动态、分布的网络环境下，分布式的语义转发与语义整合推理需要多个汇聚节点的协作和配合。首先，本文在实际网络中，模拟多个转发节点和数据源节点，接下来实现自组织语义转发和结果整合；最后，对系统的转发和整合的效率进行了评价，并与当前的研究系统进行了横向比较。本论文的研究将解决网络环境下动态分布推理和知识发现对位置的依赖性、知识发现请求与响应的时效性和准确性、内容传输的有效性等重要问题。在本论文提出的技术和算法的基础上，我们提出了基于命名数据网络的分布式推理框架、实现流程和实现模型，解决了动态分布知识的智能发现和整合的问题，为网络环境下大规模知识的发现、使用和管理提供了较好的理论和技术基础。
[Abstract]:With the development of network and communication technology, The Internet has evolved into a pervasive information infrastructure for content distribution and service delivery. Finding knowledge and rules intelligently in large-scale distributed and dynamic data sources is a challenging issue. Data-oriented, content-or information-centric network protocol architecture has transformed from "machine interconnection" to "information interconnection". In view of the huge position independent data namespace, this paper mainly studies the distribution, the self-organizing semantic reasoning method, the algorithm model and the system implementation method. First, a user-driven semantic reasoning mechanism suitable for naming data network environment is studied. Naming data network is to separate the name of content from its location. In the dynamically distributed network environment, in order to discover all the potential data sources related to the user's needs, the semantic forwarding process of user inference requests is studied. A forwarding engine with reasoning ability is designed in forwarding node. According to the return and convergence process of distributed reasoning results, the semantic integration and aggregation mechanism and implementation method of reasoning results are studied. (3) the whole reasoning process needs to be distributed. Therefore, a dynamic adaptive and efficient reasoning mechanism and algorithm are also proposed in this paper. Secondly, the reasoning rules based on request-driven and routing process are studied. Semantic reasoning needs to be simple and effective in the dynamic distributed network environment. Then, based on the characteristics of hierarchical content naming mechanism and routing index, a request forwarding reasoning mechanism is proposed to discover all potential knowledge sources and forward requests to them. Based on the aggregate return results at the convergence point, the semantic integration rules of the aggregation results are studied to achieve the goal of refining and integrating the inference results, and the evolution rules of the inference requests are studied to update the requests according to some acquired knowledge. To discover the user needs more potential knowledge. Thirdly, the system implementation method of dynamically distributed semantic forwarding and integrated reasoning based on named data network is studied. Distributed semantic forwarding and semantic integration reasoning need the cooperation and cooperation of multiple convergent nodes. Firstly, this paper simulates multiple forwarding nodes and data source nodes in the actual network, and then realizes self-organizing semantic forwarding and result integration. Finally, the efficiency of system forwarding and integration is evaluated and compared horizontally with the current research system. The research in this paper will solve the dynamic distributed reasoning and knowledge discovery dependence on the location, the timeliness and accuracy of the request and response of knowledge discovery. Based on the techniques and algorithms proposed in this paper, we propose a distributed reasoning framework based on named data network, implementation process and implementation model. It solves the problem of intelligent discovery and integration of dynamically distributed knowledge, and provides a good theoretical and technical basis for the discovery, use and management of large-scale knowledge in the network environment.
【学位授予单位】：湖南科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP391.1;TP393.09

【参考文献】