Over-the-air computation(AirComp)based federated learning(FL)has been a promising technique for distilling artificial intelligence(AI)at the network edge.However,the performance of AirComp-based FL is decided by the d...Over-the-air computation(AirComp)based federated learning(FL)has been a promising technique for distilling artificial intelligence(AI)at the network edge.However,the performance of AirComp-based FL is decided by the device with the lowest channel gain due to the signal alignment property.More importantly,most existing work focuses on a single-cell scenario,where inter-cell interference is ignored.To overcome these shortages,a reconfigurable intelligent surface(RIS)-assisted AirComp-based FL system is proposed for multi-cell networks,where a RIS is used for enhancing the poor user signal caused by channel fading,especially for the device at the cell edge,and reducing inter-cell interference.The convergence of FL in the proposed system is first analyzed and the optimality gap for FL is derived.To minimize the optimality gap,we formulate a joint uplink and downlink optimization problem.The formulated problem is then divided into two separable nonconvex subproblems.Following the successive convex approximation(SCA)method,we first approximate the nonconvex term to a linear form,and then alternately optimize the beamforming vector and phase-shift matrix for each cell.Simulation results demonstrate the advantages of deploying a RIS in multi-cell networks and our proposed system significantly improves the performance of FL.展开更多
In this paper,we propose a reconfigurable intelligent surface(RIS)assisted over-the-air federated learning(FL),where multiple antennas are deployed at each edge device to enable simultaneous multidimensional model tra...In this paper,we propose a reconfigurable intelligent surface(RIS)assisted over-the-air federated learning(FL),where multiple antennas are deployed at each edge device to enable simultaneous multidimensional model transmission over a millimeter wave(mmWave)network.We conduct rigorous convergence analysis for the proposed FL system,taking into account dynamic channel fading and analog transmissions.Inspired by the convergence analysis,we propose to jointly optimize the receive digital and analog beamforming matrices at the access point,the RIS phase-shift matrix,as well as the transmit beamforming matrices at transmitting devices to minimize the transmission distortion.The optimization variable coupling and non-convex constraints make the formulated problem challenging to be solved.To this end,we develop a low-complexity Riemannian conjugate gradient(RCG)-based algorithm to solve the unit modulus constraints and decouple the optimization variables.Simulations show that the proposed RCG algorithm outperforms the successive convex approximation algorithm in terms of the learning performance.展开更多
文摘Over-the-air computation(AirComp)based federated learning(FL)has been a promising technique for distilling artificial intelligence(AI)at the network edge.However,the performance of AirComp-based FL is decided by the device with the lowest channel gain due to the signal alignment property.More importantly,most existing work focuses on a single-cell scenario,where inter-cell interference is ignored.To overcome these shortages,a reconfigurable intelligent surface(RIS)-assisted AirComp-based FL system is proposed for multi-cell networks,where a RIS is used for enhancing the poor user signal caused by channel fading,especially for the device at the cell edge,and reducing inter-cell interference.The convergence of FL in the proposed system is first analyzed and the optimality gap for FL is derived.To minimize the optimality gap,we formulate a joint uplink and downlink optimization problem.The formulated problem is then divided into two separable nonconvex subproblems.Following the successive convex approximation(SCA)method,we first approximate the nonconvex term to a linear form,and then alternately optimize the beamforming vector and phase-shift matrix for each cell.Simulation results demonstrate the advantages of deploying a RIS in multi-cell networks and our proposed system significantly improves the performance of FL.
基金National Natu-ral Science Foundation of China(NSFC)(62001294)IEEE Vehicular Technology Confer-ence(VTC2022-Spring),Helsinki,Finland,Jun.2022。
文摘In this paper,we propose a reconfigurable intelligent surface(RIS)assisted over-the-air federated learning(FL),where multiple antennas are deployed at each edge device to enable simultaneous multidimensional model transmission over a millimeter wave(mmWave)network.We conduct rigorous convergence analysis for the proposed FL system,taking into account dynamic channel fading and analog transmissions.Inspired by the convergence analysis,we propose to jointly optimize the receive digital and analog beamforming matrices at the access point,the RIS phase-shift matrix,as well as the transmit beamforming matrices at transmitting devices to minimize the transmission distortion.The optimization variable coupling and non-convex constraints make the formulated problem challenging to be solved.To this end,we develop a low-complexity Riemannian conjugate gradient(RCG)-based algorithm to solve the unit modulus constraints and decouple the optimization variables.Simulations show that the proposed RCG algorithm outperforms the successive convex approximation algorithm in terms of the learning performance.
