Federated learning effectively addresses issues such as data privacy by collaborating across participating devices to train global models.However,factors such as network topology and computing power of devices can aff...Federated learning effectively addresses issues such as data privacy by collaborating across participating devices to train global models.However,factors such as network topology and computing power of devices can affect its training or communication process in complex network environments.Computing and network convergence(CNC)of sixth-generation(6G)networks,a new network architecture and paradigm with computing-measurable,perceptible,distributable,dispatchable,and manageable capabilities,can effectively support federated learning training and improve its communication efficiency.By guiding the participating devices'training in federated learning based on business requirements,resource load,network conditions,and computing power of devices,CNC can reach this goal.In this paper,to improve the communication eficiency of federated learning in complex networks,we study the communication eficiency optimization methods of federated learning for CNC of 6G networks that give decisions on the training process for different network conditions and computing power of participating devices.The simulations address two architectures that exist for devices in federated learning and arrange devices to participate in training based on arithmetic power while achieving optimization of communication efficiency in the process of transferring model parameters.The results show that the methods we proposed can cope well with complex network situations,effectively balance the delay distribution of participating devices for local training,improve the communication eficiency during the transfer of model parameters,and improve the resource utilization in the network.展开更多
In this paper, we propose an energy-efficient power control scheme for device-to-device(D2D) communications underlaying cellular networks, where multiple D2D pairs reuse the same resource blocks allocated to one cellu...In this paper, we propose an energy-efficient power control scheme for device-to-device(D2D) communications underlaying cellular networks, where multiple D2D pairs reuse the same resource blocks allocated to one cellular user. Taking the maximum allowed transmit power and the minimum data rate requirement into consideration, we formulate the energy efficiency maximization problem as a non-concave fractional programming(FP) problem and then develop a two-loop iterative algorithm to solve it. In the outer loop, we adopt Dinkelbach method to equivalently transform the FP problem into a series of parametric subtractive-form problems, and in the inner loop we solve the parametric subtractive problems based on successive convex approximation and geometric programming method to obtain the solutions satisfying the KarushKuhn-Tucker conditions. Simulation results demonstrate the validity and efficiency of the proposed scheme, and illustrate the impact of different parameters on system performance.展开更多
Given the fast growth of intelligent devices, it is expected that a large number of high-stakes artificial intelligence (AI) applications, e. g., drones, autonomous cars, and tac?tile robots, will be deployed at the e...Given the fast growth of intelligent devices, it is expected that a large number of high-stakes artificial intelligence (AI) applications, e. g., drones, autonomous cars, and tac?tile robots, will be deployed at the edge of wireless networks in the near future. Therefore, the intelligent communication networks will be designed to leverage advanced wireless tech?niques and edge computing technologies to support AI-enabled applications at various end devices with limited communication, computation, hardware and energy resources. In this article, we present the principles of efficient deployment of model inference at network edge to provide low-latency and energy-efficient AI services. This includes the wireless distribut?ed computing framework for low-latency device distributed model inference as well as the wireless cooperative transmission strategy for energy-efficient edge cooperative model infer?ence. The communication efficiency of edge inference systems is further improved by build?ing up a smart radio propagation environment via intelligent reflecting surface.展开更多
The mainstream approaches to green networking are discussed first from the view of engineering,including resource consolidation,server virtualization,selective connectedness,and proportional computing.A brief introduc...The mainstream approaches to green networking are discussed first from the view of engineering,including resource consolidation,server virtualization,selective connectedness,and proportional computing.A brief introduction to network virtualization techniques is given then and a virtual node embedding approach is provided.Finally,three kinds of enhanced green networking schemes by network virtualization are proposed,that is enhancement to sever virtualization,resource consolidation and Adaptive Link Rate(ALR).Examples are included to show the virtue of network virtualization to green networking in terms of energy efficient communications.展开更多
Recent years have witnessed a spurt of progress in federated learning,which can coordinate multi-participation model training while protecting the data privacy of participants.However,low communication efficiency is a...Recent years have witnessed a spurt of progress in federated learning,which can coordinate multi-participation model training while protecting the data privacy of participants.However,low communication efficiency is a bottleneck when deploying federated learning to edge computing and IoT devices due to the need to transmit a huge number of parameters during co-training.In this paper,we verify that the outputs of the last hidden layer can record the characteristics of training data.Accordingly,we propose a communication-efficient strategy based on model split and representation aggregate.Specifically,we make the client upload the outputs of the last hidden layer instead of all model parameters when participating in the aggregation,and the server distributes gradients according to the global information to revise local models.Empirical evidence from experiments verifies that our method can complete training by uploading less than one-tenth of model parameters,while preserving the usability of the model.展开更多
Since the data samples on client devices are usually non-independent and non-identically distributed(non-IID),this will challenge the convergence of federated learning(FL)and reduce communication efficiency.This paper...Since the data samples on client devices are usually non-independent and non-identically distributed(non-IID),this will challenge the convergence of federated learning(FL)and reduce communication efficiency.This paper proposes FedQMIX,a node selection algorithm based on multi-agent reinforcement learning(MARL),to address these challenges.Firstly,we observe a connection between model weights and data distribution,and a clustering algorithm can group clients with similar data distribution into the same cluster.Secondly,we propose a QMIX-based mechanism that learns to select devices from clustering results in each communication round to maximize the reward,penalizing the use of more communication rounds and thereby improving the communication efficiency of FL.Finally,experiments show that FedQMIX can reduce the number of communication rounds by 11%and 30%on the MNIST and CIFAR-10 datasets,respectively,compared to the baseline algorithm(Favor).展开更多
In this paper, the general efficiency, which is the average of the global efficiency and the local efficiency, is defined to measure the communication efficiency of a network. The increasing ratio of the general effic...In this paper, the general efficiency, which is the average of the global efficiency and the local efficiency, is defined to measure the communication efficiency of a network. The increasing ratio of the general efficiency of a small-world network relative to that of the corresponding regular network is used to measure the small-world effect quantitatively. The more considerable the small-world effect, the higher the general efficiency of a network with a certain cost is. It is shown that the small-world effect increases monotonically with the increase of the vertex number. The optimal rewiring probability to induce the best small-world effect is approximately 0.02 and the optimal average connection probability decreases monotonically with the increase of the vertex number. Therefore, the optimal network structure to induce the maximal small-world effect is the structure with the large vertex number （〉 500）, the small rewiring probability （≈0.02） and the small average connection probability （〈 0.1）. Many previous research results support our results.展开更多
The burgeoning advances in machine learning and wireless technologies are forg?ing a new paradigm for future networks, which are expected to possess higher degrees of in?telligence via the inference from vast dataset ...The burgeoning advances in machine learning and wireless technologies are forg?ing a new paradigm for future networks, which are expected to possess higher degrees of in?telligence via the inference from vast dataset and being able to respond to local events in a timely manner. Due to the sheer volume of data generated by end-user devices, as well as the increasing concerns about sharing private information, a new branch of machine learn?ing models, namely federated learning, has emerged from the intersection of artificial intelli?gence and edge computing. In contrast to conventional machine learning methods, federated learning brings the models directly to the device for training, where only the resultant param?eters shall be sent to the edge servers. The local copies of the model on the devices bring along great advantages of eliminating network latency and preserving data privacy. Never?theless, to make federated learning possible, one needs to tackle new challenges that require a fundamental departure from standard methods designed for distributed optimizations. In this paper, we aim to deliver a comprehensive introduction of federated learning. Specifical?ly, we first survey the basis of federated learning, including its learning structure and the distinct features from conventional machine learning models. We then enumerate several critical issues associated with the deployment of federated learning in a wireless network, and show why and how technologies should be jointly integrated to facilitate the full imple?mentation from different perspectives, ranging from algorithmic design, on-device training, to communication resource management. Finally, we conclude by shedding light on some po?tential applications and future trends.展开更多
Two end-users which have symmetric traffic requirements in terms of data rate are considered. They exchange information in Rayleigh flat-fading channels and multiple serial half-duplex relay nodes are employed to exte...Two end-users which have symmetric traffic requirements in terms of data rate are considered. They exchange information in Rayleigh flat-fading channels and multiple serial half-duplex relay nodes are employed to extend the communication coverage and assist the bidirectional communication between them using the analog network coding( ANC) protocol. With the objective of minimizing the sum transmit energy at the required data rate c,the optimal relay positioning and power allocation problem is firstly investigated and then the sub-optimal solutions for a two-relay channel are proposed,due to no close-form optimal solution. Furthermore,a sub-optimal scheme of relay positioning and power allocation,called equal-distance equal-transmit-power( EDEP) for an arbitrary Nrelay channel,N > 1 is proposed. Simulation results demonstrate a consistence with our proposed scheme.展开更多
Massive MIMO is one of tile enabling technologies tbr beyond 4G and 5G systems due to its ability to provide beamforming gain and reduce interference Dual-polarized antenna is widely adopted to accommodate a large num...Massive MIMO is one of tile enabling technologies tbr beyond 4G and 5G systems due to its ability to provide beamforming gain and reduce interference Dual-polarized antenna is widely adopted to accommodate a large number of antenna elements in limited space. However, current CSI（channel state information） feedback schemes developed in LTE for conventional MIMO systems are not efficient enough for massive MIMO systems since the overhead increases almost linearly with the number of antenna. Moreover, the codebook for massive MIMO will be huge and difficult to design with the LTE methodology. This paper proposes a novel CSI feedback scheme named layered Multi-paths Information based CSI Feedback （LMPIF）, which can achieve higher spectrum efficiency for dual-polarized antenna system with low feedback overhead. The MIMO channel is decomposed into long term components （multipath directions and amplitudes） and short term components （multipath phases）. The relationship between the two components and the optimal precoder is derived in closed form. To reduce the overhead, different granularities in feedback time have been applied for the long term components and short term components Link and system level simulation results prove that LMPIF can improve performance considerably with low CSI feedback overhead.展开更多
"Device-independent"not only represents a relaxation of the security assumptions about the internal working of the quantum devices,but also can enhance the security of the quantum communication.In the paper,..."Device-independent"not only represents a relaxation of the security assumptions about the internal working of the quantum devices,but also can enhance the security of the quantum communication.In the paper,we put forward the first device-independent quantum secure direct communication(DIQSDC)protocol and analyze its security and communication efficiency against collective attacks.Under practical noisy quantum channel condition,the photon transmission loss and photon state decoherence would reduce DI-QSDC’s communication quality and threaten its absolute security.For solving the photon transmission loss and decoherence problems,we adopt noiseless linear amplification(NLA)protocol and entanglement purification protocol(EPP)to modify the DI-QSDC protocol.With the help of the NLA and EPP,we can guarantee DI-QSDC’s absolute security and effectively improve its communication quality.展开更多
In order to make full use of the radio resource of heterogeneous wireless networks（HWNs） and promote the quality of service（Qo S） of multi-homing users for video communication, a bandwidth allocation algorithm bas...In order to make full use of the radio resource of heterogeneous wireless networks（HWNs） and promote the quality of service（Qo S） of multi-homing users for video communication, a bandwidth allocation algorithm based on multi-radio access is proposed in this paper. The proposed algorithm adopts an improved distributed common radio resource management（DCRRM） model which can reduce the signaling overhead sufficiently. This scheme can be divided into two phases. In the first phase, candidate network set of each user is obtained according to the received signal strength（RSS）. And the simple additive weighted（SAW） method is employed to determine the active network set. In the second phase, the utility optimization problem is formulated by linear combining of the video communication satisfaction model, cost model and energy efficiency model. And finding the optimal bandwidth allocation scheme with Lagrange multiplier method and Karush-Kuhn-Tucker（KKT） conditions. Simulation results show that the proposed algorithm promotes the network load performances and guarantees that users obtain the best joint utility under current situation.展开更多
In this paper,distributed estimation of high-dimensional sparse precision matrix is proposed based on the debiased D-trace loss penalized lasso and the hard threshold method when samples are distributed into different...In this paper,distributed estimation of high-dimensional sparse precision matrix is proposed based on the debiased D-trace loss penalized lasso and the hard threshold method when samples are distributed into different machines for transelliptical graphical models.At a certain level of sparseness,this method not only achieves the correct selection of non-zero elements of sparse precision matrix,but the error rate can be comparable to the estimator in a non-distributed setting.The numerical results further prove that the proposed distributed method is more effective than the usual average method.展开更多
基金supported by the National Natural Science Foundation of China(Nos.62271062 and 62071063)。
文摘Federated learning effectively addresses issues such as data privacy by collaborating across participating devices to train global models.However,factors such as network topology and computing power of devices can affect its training or communication process in complex network environments.Computing and network convergence(CNC)of sixth-generation(6G)networks,a new network architecture and paradigm with computing-measurable,perceptible,distributable,dispatchable,and manageable capabilities,can effectively support federated learning training and improve its communication efficiency.By guiding the participating devices'training in federated learning based on business requirements,resource load,network conditions,and computing power of devices,CNC can reach this goal.In this paper,to improve the communication eficiency of federated learning in complex networks,we study the communication eficiency optimization methods of federated learning for CNC of 6G networks that give decisions on the training process for different network conditions and computing power of participating devices.The simulations address two architectures that exist for devices in federated learning and arrange devices to participate in training based on arithmetic power while achieving optimization of communication efficiency in the process of transferring model parameters.The results show that the methods we proposed can cope well with complex network situations,effectively balance the delay distribution of participating devices for local training,improve the communication eficiency during the transfer of model parameters,and improve the resource utilization in the network.
基金supported by National Natural Science Foundation of China (No.61501028)Beijing Institute of Technology Research Fund Program for Young Scholars
文摘In this paper, we propose an energy-efficient power control scheme for device-to-device(D2D) communications underlaying cellular networks, where multiple D2D pairs reuse the same resource blocks allocated to one cellular user. Taking the maximum allowed transmit power and the minimum data rate requirement into consideration, we formulate the energy efficiency maximization problem as a non-concave fractional programming(FP) problem and then develop a two-loop iterative algorithm to solve it. In the outer loop, we adopt Dinkelbach method to equivalently transform the FP problem into a series of parametric subtractive-form problems, and in the inner loop we solve the parametric subtractive problems based on successive convex approximation and geometric programming method to obtain the solutions satisfying the KarushKuhn-Tucker conditions. Simulation results demonstrate the validity and efficiency of the proposed scheme, and illustrate the impact of different parameters on system performance.
文摘Given the fast growth of intelligent devices, it is expected that a large number of high-stakes artificial intelligence (AI) applications, e. g., drones, autonomous cars, and tac?tile robots, will be deployed at the edge of wireless networks in the near future. Therefore, the intelligent communication networks will be designed to leverage advanced wireless tech?niques and edge computing technologies to support AI-enabled applications at various end devices with limited communication, computation, hardware and energy resources. In this article, we present the principles of efficient deployment of model inference at network edge to provide low-latency and energy-efficient AI services. This includes the wireless distribut?ed computing framework for low-latency device distributed model inference as well as the wireless cooperative transmission strategy for energy-efficient edge cooperative model infer?ence. The communication efficiency of edge inference systems is further improved by build?ing up a smart radio propagation environment via intelligent reflecting surface.
基金the National Natural Science Foundation of China，the PAPD Project of Jiangsu Higher Education Institutions，the National S&T Dedicated Mega-Project，the Qing Lan Project of Jiangsu Province of China，the open research fund of Key Lab of Broadband Wireless Communication and Sensor Network Technology (Nanjing University of Posts and Telecommunications),Ministry of Education
文摘The mainstream approaches to green networking are discussed first from the view of engineering,including resource consolidation,server virtualization,selective connectedness,and proportional computing.A brief introduction to network virtualization techniques is given then and a virtual node embedding approach is provided.Finally,three kinds of enhanced green networking schemes by network virtualization are proposed,that is enhancement to sever virtualization,resource consolidation and Adaptive Link Rate(ALR).Examples are included to show the virtue of network virtualization to green networking in terms of energy efficient communications.
基金supported by Shenzhen Basic Research (General Project)under Grant No.JCYJ20190806142601687Shenzhen Stable Supporting Program (General Project) under Grant No.GXWD20201230155427003-20200821160539001+1 种基金Guangdong Provincial Key Laboratory of Novel Security Intelligence Technologies under Grant No.2022B1212010005Shenzhen Basic Research (Key Project) under Grant No.JCYJ20200109113405927。
文摘Recent years have witnessed a spurt of progress in federated learning,which can coordinate multi-participation model training while protecting the data privacy of participants.However,low communication efficiency is a bottleneck when deploying federated learning to edge computing and IoT devices due to the need to transmit a huge number of parameters during co-training.In this paper,we verify that the outputs of the last hidden layer can record the characteristics of training data.Accordingly,we propose a communication-efficient strategy based on model split and representation aggregate.Specifically,we make the client upload the outputs of the last hidden layer instead of all model parameters when participating in the aggregation,and the server distributes gradients according to the global information to revise local models.Empirical evidence from experiments verifies that our method can complete training by uploading less than one-tenth of model parameters,while preserving the usability of the model.
基金supported by the National Natural Science Foundation of China(NSFC)(62072469)。
文摘Since the data samples on client devices are usually non-independent and non-identically distributed(non-IID),this will challenge the convergence of federated learning(FL)and reduce communication efficiency.This paper proposes FedQMIX,a node selection algorithm based on multi-agent reinforcement learning(MARL),to address these challenges.Firstly,we observe a connection between model weights and data distribution,and a clustering algorithm can group clients with similar data distribution into the same cluster.Secondly,we propose a QMIX-based mechanism that learns to select devices from clustering results in each communication round to maximize the reward,penalizing the use of more communication rounds and thereby improving the communication efficiency of FL.Finally,experiments show that FedQMIX can reduce the number of communication rounds by 11%and 30%on the MNIST and CIFAR-10 datasets,respectively,compared to the baseline algorithm(Favor).
基金Project supported by the National Natural Science Foundation of China (Grant Nos.61101117,61171099,and 61362008)the National Key Scientific and Technological Project of China (Grant No.2012ZX03004005002)+1 种基金the Fundamental Research Funds for the Central Universities,China (Grant No.BUPT2012RC0112)the Natural Science Foundation of Jiangxi Province,China (Grant No.20132BAB201018)
文摘In this paper, the general efficiency, which is the average of the global efficiency and the local efficiency, is defined to measure the communication efficiency of a network. The increasing ratio of the general efficiency of a small-world network relative to that of the corresponding regular network is used to measure the small-world effect quantitatively. The more considerable the small-world effect, the higher the general efficiency of a network with a certain cost is. It is shown that the small-world effect increases monotonically with the increase of the vertex number. The optimal rewiring probability to induce the best small-world effect is approximately 0.02 and the optimal average connection probability decreases monotonically with the increase of the vertex number. Therefore, the optimal network structure to induce the maximal small-world effect is the structure with the large vertex number （〉 500）, the small rewiring probability （≈0.02） and the small average connection probability （〈 0.1）. Many previous research results support our results.
文摘The burgeoning advances in machine learning and wireless technologies are forg?ing a new paradigm for future networks, which are expected to possess higher degrees of in?telligence via the inference from vast dataset and being able to respond to local events in a timely manner. Due to the sheer volume of data generated by end-user devices, as well as the increasing concerns about sharing private information, a new branch of machine learn?ing models, namely federated learning, has emerged from the intersection of artificial intelli?gence and edge computing. In contrast to conventional machine learning methods, federated learning brings the models directly to the device for training, where only the resultant param?eters shall be sent to the edge servers. The local copies of the model on the devices bring along great advantages of eliminating network latency and preserving data privacy. Never?theless, to make federated learning possible, one needs to tackle new challenges that require a fundamental departure from standard methods designed for distributed optimizations. In this paper, we aim to deliver a comprehensive introduction of federated learning. Specifical?ly, we first survey the basis of federated learning, including its learning structure and the distinct features from conventional machine learning models. We then enumerate several critical issues associated with the deployment of federated learning in a wireless network, and show why and how technologies should be jointly integrated to facilitate the full imple?mentation from different perspectives, ranging from algorithmic design, on-device training, to communication resource management. Finally, we conclude by shedding light on some po?tential applications and future trends.
基金National Natural Science Foundation of China(No.61071214)
文摘Two end-users which have symmetric traffic requirements in terms of data rate are considered. They exchange information in Rayleigh flat-fading channels and multiple serial half-duplex relay nodes are employed to extend the communication coverage and assist the bidirectional communication between them using the analog network coding( ANC) protocol. With the objective of minimizing the sum transmit energy at the required data rate c,the optimal relay positioning and power allocation problem is firstly investigated and then the sub-optimal solutions for a two-relay channel are proposed,due to no close-form optimal solution. Furthermore,a sub-optimal scheme of relay positioning and power allocation,called equal-distance equal-transmit-power( EDEP) for an arbitrary Nrelay channel,N > 1 is proposed. Simulation results demonstrate a consistence with our proposed scheme.
基金supported by the National High-Tech R&D Program(863 Program 2015AA01A705)
文摘Massive MIMO is one of tile enabling technologies tbr beyond 4G and 5G systems due to its ability to provide beamforming gain and reduce interference Dual-polarized antenna is widely adopted to accommodate a large number of antenna elements in limited space. However, current CSI（channel state information） feedback schemes developed in LTE for conventional MIMO systems are not efficient enough for massive MIMO systems since the overhead increases almost linearly with the number of antenna. Moreover, the codebook for massive MIMO will be huge and difficult to design with the LTE methodology. This paper proposes a novel CSI feedback scheme named layered Multi-paths Information based CSI Feedback （LMPIF）, which can achieve higher spectrum efficiency for dual-polarized antenna system with low feedback overhead. The MIMO channel is decomposed into long term components （multipath directions and amplitudes） and short term components （multipath phases）. The relationship between the two components and the optimal precoder is derived in closed form. To reduce the overhead, different granularities in feedback time have been applied for the long term components and short term components Link and system level simulation results prove that LMPIF can improve performance considerably with low CSI feedback overhead.
基金supported by the National Natural Science Foundation of China (11974189 and 11974205)the China Postdoctoral Science Foundation (2018M642293)+1 种基金the Open Research Fund of the Key Lab of Broadband Wireless Communication and Sensor Network Technology,Nanjing University of Posts and Telecommunications, Ministry of Education (JZNY201908)a Project Funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions
文摘"Device-independent"not only represents a relaxation of the security assumptions about the internal working of the quantum devices,but also can enhance the security of the quantum communication.In the paper,we put forward the first device-independent quantum secure direct communication(DIQSDC)protocol and analyze its security and communication efficiency against collective attacks.Under practical noisy quantum channel condition,the photon transmission loss and photon state decoherence would reduce DI-QSDC’s communication quality and threaten its absolute security.For solving the photon transmission loss and decoherence problems,we adopt noiseless linear amplification(NLA)protocol and entanglement purification protocol(EPP)to modify the DI-QSDC protocol.With the help of the NLA and EPP,we can guarantee DI-QSDC’s absolute security and effectively improve its communication quality.
基金supported by the National Natural Science Foundation of China (61571234, 61401225)the National Basic Research Program of China (2013CB329005)+1 种基金the Hi-Tech Research and Development Program of China (2014AA01A705)the Graduate Student Innovation Plan of Jiangsu Province (SJLX15_0365)
文摘In order to make full use of the radio resource of heterogeneous wireless networks（HWNs） and promote the quality of service（Qo S） of multi-homing users for video communication, a bandwidth allocation algorithm based on multi-radio access is proposed in this paper. The proposed algorithm adopts an improved distributed common radio resource management（DCRRM） model which can reduce the signaling overhead sufficiently. This scheme can be divided into two phases. In the first phase, candidate network set of each user is obtained according to the received signal strength（RSS）. And the simple additive weighted（SAW） method is employed to determine the active network set. In the second phase, the utility optimization problem is formulated by linear combining of the video communication satisfaction model, cost model and energy efficiency model. And finding the optimal bandwidth allocation scheme with Lagrange multiplier method and Karush-Kuhn-Tucker（KKT） conditions. Simulation results show that the proposed algorithm promotes the network load performances and guarantees that users obtain the best joint utility under current situation.
基金partly supported by National Natural Science Foundation of China(Grant Nos.12031016,11971324,11471223)Foundations of Science and Technology Innovation Service Capacity Building,Interdisciplinary Construction of Bioinformatics and Statistics,and Academy for Multidisciplinary Studies,Capital Normal University,Beijing。
文摘In this paper,distributed estimation of high-dimensional sparse precision matrix is proposed based on the debiased D-trace loss penalized lasso and the hard threshold method when samples are distributed into different machines for transelliptical graphical models.At a certain level of sparseness,this method not only achieves the correct selection of non-zero elements of sparse precision matrix,but the error rate can be comparable to the estimator in a non-distributed setting.The numerical results further prove that the proposed distributed method is more effective than the usual average method.
