A hierarchical federated learning framework based on wireless device-to-device (D2D) networks was proposed to solve the problem of large communication resource consumption and limited device computing resources faced by deploying federated learning in wireless networks. Different from the traditional architectures, the hierarchical aggregation was adopted for model training. The architecture performed the intra-cluster aggregation through D2D networks, and each cluster performed the decentralized training at the same time. A cluster head was selected from each cluster to upload the model to the server for global aggregation. The network traffic of the central node was reduced by combining the hierarchical federated learning and decentralized learning. The degree of the vertices in the D2D networks was used to measure the model convergence performance. The head selection and bandwidth allocation were jointly optimized by maximizing the total degree of selected cluster heads. An optimization algorithm based on dynamic programming was designed to obtain the optimal solutions. The simulation results show that compared with the baseline algorithm,the framework can not only effectively reduce the frequency of global aggregation and training time, but also improve the performance of the final model.
Fig.1System model of hierarchical federated learning system based on wireless D2D networks
Fig.2Sequence model of hierarchical federated learning
Fig.3Local experience loss versus degree for training CNN on Cifar-10
Fig.4Local experience loss versus degree for training DNN on MNIST
Fig.5Accuracy of CNN on Cifar-10 varies with time in different algorithms
Fig.6Accuracy of DNN on MNIST varies with time in different algorithms
Fig.7Accuracy of CNN on Cifar-10 varies with time in different cluster head selection methods
Fig.8Accuracy of DNN on MNIST varies with time in different cluster head selection methods
Fig.9Accuracy of CNN on Cifar-10 varies with time in different global aggregation frequencies
Fig.10Accuracy of DNN on MNIST varies with time in different global aggregation frequencies
[1]
CHEN M, YANG Z, SAAD W, et al A Joint learning and communications framework for federated learning over wireless networks[J]. IEEE Transactions on Wireless Communications, 2020, 20 (1): 269- 283
[2]
ZHU G, LIU D, DU Y, et al Towards an intelligent edge: wireless communication meets machine learning[J]. IEEE Communications Magazine, 2020, 58 (1): 19- 25
doi: 10.1109/MCOM.001.1900103
[3]
CHEN M, GUNDUZ D, HUANG K, et al Distributed learning in wireless networks: recent progress and future challenges[J]. IEEE Journal on Selected Areas in Communications, 2021, 39 (12): 3579- 3605
doi: 10.1109/JSAC.2021.3118346
[4]
MCMAHAN B, MOORE E, RAMAGE D, et al. Communication-efficient learning of deep networks from decentralized data [C]// Artificial Intelligence and Statistics. Lauderdale: PMLR, 2017: 1273-1282.
[5]
LIM W Y B, LUONG N C, HOANG D T, et al Federated learning in mobile edge networks: a comprehensive survey[J]. IEEE Communications Surveys and Tutorials, 2020, 22 (3): 2031- 2063
doi: 10.1109/COMST.2020.2986024
[6]
LIM W Y B, NG J S, XIONG Z, et al Decentralized edge intelligence: a dynamic resource allocation framework for hierarchical federated learning[J]. IEEE Transactions on Parallel and Distributed Systems, 2021, 33 (3): 536- 550
[7]
WANG Z, XU H, LIU J, et al. Accelerating federated learning with cluster construction and hierarchical aggregation [EB/OL]. [2022-02-01]. https://www.computer.org/csdl/journal/tm/5555/01/09699080/1ADJeS80Gvm.
[8]
FENG D, LU L, YUANWU Y, et al Device-to-device communications in cellular networks[J]. IEEE Communications Magazine, 2014, 52 (4): 49- 55
doi: 10.1109/MCOM.2014.6807946
[9]
ASADI A, WANG Q, MANCUSO V A survey on device-to-device communication in cellular networks[J]. IEEE Communications Surveys and Tutorials, 2014, 16 (4): 1801- 1819
doi: 10.1109/COMST.2014.2319555
[10]
YU D, ZOU Z, CHEN S, et al Decentralized parallel sgd with privacy preservation in vehicular networks[J]. IEEE Transactions on Vehicular Technology, 2021, 70 (6): 5211- 5220
doi: 10.1109/TVT.2021.3064877
[11]
XING H, SIMEONE O, BI S. Decentralized federated learning via SGD over wireless D2D networks [C]// 21st International Workshop on Signal Processing Advances in Wireless Communications. Atlanta: IEEE, 2020: 1-5.
[12]
POKHREL S R, CHOI J. A decentralized federated learning approach for connected autonomous vehicles [C]// IEEE Wireless Communications and Networking Conference Workshops. Seoul : IEEE, 2020: 1-6.
[13]
ZHU L, LIU C, YUAN J, et al. Machine learning-based resource optimization for D2D communication underlaying networks [C]// IEEE 92nd Vehicular Technology Conference (VTC2020-Fall). Victoria: IEEE, 2020: 1-6.
[14]
HOANG T D, LE L B, LENGOC T Resource allocation for D2D communication underlaid cellular networks using graph-based approach[J]. IEEE Transactions on Wireless Communications, 2016, 15 (10): 7099- 7113
doi: 10.1109/TWC.2016.2597283
[15]
OZFATURA E, RINI S, GÜNDÜZ D. Decentralized SGD with over-the-air computation [C]// IEEE Global Communications Conference. Taipei: IEEE, 2020: 1-6.
[16]
NISHIO T, YONETANI R. Client selection for federated learning with heterogeneous resources in mobile edge [C]// IEEE International Conference on Communications. Shanghai: IEEE, 2019: 1-7.
[17]
WEN D, BENNIS M, HUANG K Joint parameter-and-bandwidth allocation for improving the efficiency of partitioned edge learning[J]. IEEE Transactions on Wireless Communications, 2020, 19 (12): 8272- 8286
doi: 10.1109/TWC.2020.3021177
[18]
JIANG Z, YU G, CAI Y, Decentralized edge learning via unreliable device-to-device communications [J]. IEEE Transactions on Wireless Communications, 2022, 21(11): 9041-9055
[19]
WEN D, JEON K J, BENNIS M, et al Adaptive subcarrier, parameter, and power allocation for partitioned edge learning over broadband channels[J]. IEEE Transactions on Wireless Communications, 2021, 20 (12): 8348- 8361
doi: 10.1109/TWC.2021.3092075
[20]
ZENG Q, DU Y, HUANG K, et al Energy-efficient resource management for federated edge learning with CPU-GPU heterogeneous computing[J]. IEEE Transactions on Wireless Communications, 2021, 20 (12): 7947- 7962
doi: 10.1109/TWC.2021.3088910
[21]
KOLOSKOVA A, LOIZOU N, BOREIRI S, et al. A unified theory of decentralized sgd with changing topology and local updates [C]// International Conference on Machine Learning. Vienna: PMLR, 2020: 5381-5393.
