Joint Device Scheduling and Resource Allocation for Latency Constrained Wireless Federated Learning

  • 2020-07-14 16:46:47
  • Wenqi Shi, Sheng Zhou, Zhisheng Niu, Miao Jiang, Lu Geng
  • 0

Abstract

In federated learning (FL), devices contribute to the global training byuploading their local model updates via wireless channels. Due to limitedcomputation and communication resources, device scheduling is crucial to theconvergence rate of FL. In this paper, we propose a joint device scheduling andresource allocation policy to maximize the model accuracy within a given totaltraining time budget for latency constrained wireless FL. A lower bound on thereciprocal of the training performance loss, in terms of the number of trainingrounds and the number of scheduled devices per round, is derived. Based on thebound, the accuracy maximization problem is solved by decoupling it into twosub-problems. First, given the scheduled devices, the optimal bandwidthallocation suggests allocating more bandwidth to the devices with worse channelconditions or weaker computation capabilities. Then, a greedy device schedulingalgorithm is introduced, which in each step selects the device consuming theleast updating time obtained by the optimal bandwidth allocation, until thelower bound begins to increase, meaning that scheduling more devices willdegrade the model accuracy. Experiments show that the proposed policyoutperforms state-of-the-art scheduling policies under extensive settings ofdata distributions and cell radius.

 

Quick Read (beta)

loading the full paper ...