Coded Matrix Computations for D2D-enabled Linearized Federated Learning
This work addresses straggler and privacy issues in federated learning for distributed client devices, representing an incremental improvement over existing D2D methods.
The paper tackles the problem of straggler clients and privacy concerns in device-to-device (D2D) enabled federated learning by proposing a coded matrix computations approach, resulting in reduced communication delay and improved local computation speed for sparse data matrices, as confirmed by numerical evaluations.
Federated learning (FL) is a popular technique for training a global model on data distributed across client devices. Like other distributed training techniques, FL is susceptible to straggler (slower or failed) clients. Recent work has proposed to address this through device-to-device (D2D) offloading, which introduces privacy concerns. In this paper, we propose a novel straggler-optimal approach for coded matrix computations which can significantly reduce the communication delay and privacy issues introduced from D2D data transmissions in FL. Moreover, our proposed approach leads to a considerable improvement of the local computation speed when the generated data matrix is sparse. Numerical evaluations confirm the superiority of our proposed method over baseline approaches.