5G is envisioned to guarantee high transmission rate,ultra-low latency,high reliability and massive connections.To satisfy the above requirements,the 5G architecture is designed with the properties of using service-ba...5G is envisioned to guarantee high transmission rate,ultra-low latency,high reliability and massive connections.To satisfy the above requirements,the 5G architecture is designed with the properties of using service-based architecture,cloud-native oriented,adopting IT-based API interfaces and introduction of the Network Repository Function.However,with the wide commercialization of 5G network and the exploration towards 6G,the 5G architecture exposes the disadvantages of high architecture complexity,difficult inter-interface communication,low cognitive capability,bad instantaneity,and deficient intelligence.To overcome these limitations,this paper investigates 6G network architecture,and proposes a cognitive intelligence based distributed 6G network architecture.This architecture consists of a physical network layer and an intelligent decision layer.The two layers coordinate through flexible service interfaces,supporting function decoupling and joint evolution of intelligence services and network services.With the above design,the proposed 6G architecture can be updated autonomously to deal with the future unpredicted complex services.展开更多
Having criticized the current architecture of Advanced Traveler Information Systems （ATISs）, this work discusses a new base of requirements to develop a new paradigm for traffic information systems. It mainly integr...Having criticized the current architecture of Advanced Traveler Information Systems （ATISs）, this work discusses a new base of requirements to develop a new paradigm for traffic information systems. It mainly integrates three dimensions within a traffic system： drivers＇ pattern of behavior and preferences, urban traffic desires, and capabilities of traffic information service providers. Based on the above, functional segments from several related backgrounds are brought together to structure a new architecture, called Interactive Traveler Information System （ITIS）. The main interactive feature of this new architecture is a two-way communication track between drivers and the traffic information system provider-in fact, a decision on choosing a road at a particular time for an individual will be made based on the utility of both sides. This new configuration consists of driver-side smartphone application, centric traffic prediction, and decision-maker units, which will shape a new approach of decision-making processes. These all work together to satisfy the designated goal of ITIS, which is preserving the Wardrop equilibrium condition in the traffic network level. Finally, we concentrate on a comparison study, which shows a differentiation between performance of the proposed ITIS and the current ATIS model in a real situation. This has been done with simulations of analogical scenarios. The most noticeable advantage of the proposed architecture is not being limited to a saturation limit, and the positive effect of increasing system penetration in the performance of the newly introduced information system. In conclusion, new research subjects are suggested to be carried out.展开更多
以小世界模型为理论基础,以 Region 为基本逻辑管理单位,按用户需求和共享目的组织 Region。提出了基于 Region 的多层结构 Peer-to-Peer 网络模型和构造规则,给出了 Region 的划分策略和数学模型,证明了模型的正确和合理性;对模型中的...以小世界模型为理论基础,以 Region 为基本逻辑管理单位,按用户需求和共享目的组织 Region。提出了基于 Region 的多层结构 Peer-to-Peer 网络模型和构造规则,给出了 Region 的划分策略和数学模型,证明了模型的正确和合理性;对模型中的层和域、中心节点、普通节点和汇聚点进行了明确的定义,给出了节点加入、离开、中心节点选取策略和算法描述;使定位某种服务的工作量和查询范围从网络中的所有结点数降低到 Region 的节点数,有效地防止了恶意请求引发的洪,网络系统开销为常数。模拟分析表明,该模型可有效解决可扩展性、性能与效率不高问题,且网络规模越大,其综合性能的优越性越明显,因此,模型是合理有效的。展开更多
Wireless networks are playing an imperative role in our daily existence;in current scenario, the users want wireless connectivity for all location with all types of service. One of the major challenges for wireless ne...Wireless networks are playing an imperative role in our daily existence;in current scenario, the users want wireless connectivity for all location with all types of service. One of the major challenges for wireless network is security issue. First and foremost task is to detect the security attacks in the network and the second task is to prevent from an authorized attacks. In our view, a lot of researches are going on and somehow we have succeeded in the first case but the second task is very tough due to wireless channel. Our research is based on how to avoid network attack i.e. misbehavior node attack in the WiMAX system. In this paper we have proposed an algorithm for WiMAX network and our algorithm are able to prevent fixed as well as mobile misbehavior node attacks. As we know, misbehavior node misbehaves in the sense that the node does not esteem its MAC protocols and avariciously sends its packets without any restriction (Flooding of packets) because it doesn't follow any protocol. Our proposed work based on the standard time required for communication for valid user with some threshold time for valid delay and some unwanted delay with network conditions. Our approach can control continuous flooding of packets and continuously transmits Constant Bit Rate (CBR) packets by misbehavior node, which introduces noise in the network and upset the performance of the network. In the mean while the valid user communicate in a trouble-free approach.展开更多
In this paper, the architecture of trustworthy and controllable networks is discussed to meet arising application requirements. After reviewing the lessons and experiences of success and failure in the Internet and su...In this paper, the architecture of trustworthy and controllable networks is discussed to meet arising application requirements. After reviewing the lessons and experiences of success and failure in the Internet and summarizing related work, we analyze the basic targets of providing trustworthiness and controllability. Then, the anticipant architecture is introduced. Based on the resulting design, several trustworthy and controllable mechanisms are also discussed.展开更多
针对深度学习应用技术进行了研究性综述。详细阐述了RBM(受限玻尔兹曼机)逐层预训练后再用BP(反向传播)微调的深度学习贪婪层训练方法,对比分析了BP算法中三种梯度下降的方式,建议在线学习系统采用随机梯度下降,静态离线学习系统采用随...针对深度学习应用技术进行了研究性综述。详细阐述了RBM(受限玻尔兹曼机)逐层预训练后再用BP(反向传播)微调的深度学习贪婪层训练方法,对比分析了BP算法中三种梯度下降的方式,建议在线学习系统采用随机梯度下降,静态离线学习系统采用随机小批量梯度下降;归纳总结了深度学习深层结构特征,并推荐了目前最受欢迎的五层深度网络结构设计方法。分析了前馈神经网络非线性激活函数的必要性及常用的激活函数优点,并推荐Re LU(rectified linear units)激活函数。最后简要概括了深度卷积神经网络、深度递归神经网络、长短期记忆网络等新型深度网络的特点及应用场景,归纳总结了当前深度学习可能的发展方向。展开更多
基金funded by Beijing University of Posts and Telecommunications-China Mobile Research Institute Joint Innovation Center,the National Key R&D Program of China(2018YFE0205503)the National Natural Science Foundation of China(61902036,62032003,61922017)Fundamental Research Funds for the Central Universities。
文摘5G is envisioned to guarantee high transmission rate,ultra-low latency,high reliability and massive connections.To satisfy the above requirements,the 5G architecture is designed with the properties of using service-based architecture,cloud-native oriented,adopting IT-based API interfaces and introduction of the Network Repository Function.However,with the wide commercialization of 5G network and the exploration towards 6G,the 5G architecture exposes the disadvantages of high architecture complexity,difficult inter-interface communication,low cognitive capability,bad instantaneity,and deficient intelligence.To overcome these limitations,this paper investigates 6G network architecture,and proposes a cognitive intelligence based distributed 6G network architecture.This architecture consists of a physical network layer and an intelligent decision layer.The two layers coordinate through flexible service interfaces,supporting function decoupling and joint evolution of intelligence services and network services.With the above design,the proposed 6G architecture can be updated autonomously to deal with the future unpredicted complex services.
文摘Having criticized the current architecture of Advanced Traveler Information Systems （ATISs）, this work discusses a new base of requirements to develop a new paradigm for traffic information systems. It mainly integrates three dimensions within a traffic system： drivers＇ pattern of behavior and preferences, urban traffic desires, and capabilities of traffic information service providers. Based on the above, functional segments from several related backgrounds are brought together to structure a new architecture, called Interactive Traveler Information System （ITIS）. The main interactive feature of this new architecture is a two-way communication track between drivers and the traffic information system provider-in fact, a decision on choosing a road at a particular time for an individual will be made based on the utility of both sides. This new configuration consists of driver-side smartphone application, centric traffic prediction, and decision-maker units, which will shape a new approach of decision-making processes. These all work together to satisfy the designated goal of ITIS, which is preserving the Wardrop equilibrium condition in the traffic network level. Finally, we concentrate on a comparison study, which shows a differentiation between performance of the proposed ITIS and the current ATIS model in a real situation. This has been done with simulations of analogical scenarios. The most noticeable advantage of the proposed architecture is not being limited to a saturation limit, and the positive effect of increasing system penetration in the performance of the newly introduced information system. In conclusion, new research subjects are suggested to be carried out.
文摘以小世界模型为理论基础,以 Region 为基本逻辑管理单位,按用户需求和共享目的组织 Region。提出了基于 Region 的多层结构 Peer-to-Peer 网络模型和构造规则,给出了 Region 的划分策略和数学模型,证明了模型的正确和合理性;对模型中的层和域、中心节点、普通节点和汇聚点进行了明确的定义,给出了节点加入、离开、中心节点选取策略和算法描述;使定位某种服务的工作量和查询范围从网络中的所有结点数降低到 Region 的节点数,有效地防止了恶意请求引发的洪,网络系统开销为常数。模拟分析表明,该模型可有效解决可扩展性、性能与效率不高问题,且网络规模越大,其综合性能的优越性越明显,因此,模型是合理有效的。
文摘Wireless networks are playing an imperative role in our daily existence;in current scenario, the users want wireless connectivity for all location with all types of service. One of the major challenges for wireless network is security issue. First and foremost task is to detect the security attacks in the network and the second task is to prevent from an authorized attacks. In our view, a lot of researches are going on and somehow we have succeeded in the first case but the second task is very tough due to wireless channel. Our research is based on how to avoid network attack i.e. misbehavior node attack in the WiMAX system. In this paper we have proposed an algorithm for WiMAX network and our algorithm are able to prevent fixed as well as mobile misbehavior node attacks. As we know, misbehavior node misbehaves in the sense that the node does not esteem its MAC protocols and avariciously sends its packets without any restriction (Flooding of packets) because it doesn't follow any protocol. Our proposed work based on the standard time required for communication for valid user with some threshold time for valid delay and some unwanted delay with network conditions. Our approach can control continuous flooding of packets and continuously transmits Constant Bit Rate (CBR) packets by misbehavior node, which introduces noise in the network and upset the performance of the network. In the mean while the valid user communicate in a trouble-free approach.
基金Supported by the National Natural Science Foundation of China under Grant Nos. 60503052, 60573144, 90412012, and the National Grand Fundamental Research 973 Program of China under Grant No. 2003CB314804.
文摘In this paper, the architecture of trustworthy and controllable networks is discussed to meet arising application requirements. After reviewing the lessons and experiences of success and failure in the Internet and summarizing related work, we analyze the basic targets of providing trustworthiness and controllability. Then, the anticipant architecture is introduced. Based on the resulting design, several trustworthy and controllable mechanisms are also discussed.
基金Supported by the National High-Tech Research and Development Plan of China under Grant No.2006AA01Z206 (国家高技术研究发展计划(863))the National Key Project of Scientific and Technical Supporting Programs of China under Grant No.2006BAH02A03 (国家"十一五"科技支撑计划)
文摘针对深度学习应用技术进行了研究性综述。详细阐述了RBM(受限玻尔兹曼机)逐层预训练后再用BP(反向传播)微调的深度学习贪婪层训练方法,对比分析了BP算法中三种梯度下降的方式,建议在线学习系统采用随机梯度下降,静态离线学习系统采用随机小批量梯度下降;归纳总结了深度学习深层结构特征,并推荐了目前最受欢迎的五层深度网络结构设计方法。分析了前馈神经网络非线性激活函数的必要性及常用的激活函数优点,并推荐Re LU(rectified linear units)激活函数。最后简要概括了深度卷积神经网络、深度递归神经网络、长短期记忆网络等新型深度网络的特点及应用场景,归纳总结了当前深度学习可能的发展方向。
