A dual approach for federated learning
This work addresses federated learning optimization, offering incremental improvements for researchers and practitioners in distributed machine learning.
The paper tackles federated optimization by proposing a dual perspective algorithm, FedDCD, enhanced with inexact gradient oracles and Nesterov's acceleration, achieving better theoretical convergence rates than state-of-the-art primal methods in certain situations, as supported by numerical experiments on real-world datasets.
We study the federated optimization problem from a dual perspective and propose a new algorithm termed federated dual coordinate descent (FedDCD), which is based on a type of coordinate descent method developed by Necora et al.[Journal of Optimization Theory and Applications, 2017]. Additionally, we enhance the FedDCD method with inexact gradient oracles and Nesterov's acceleration. We demonstrate theoretically that our proposed approach achieves better convergence rates than the state-of-the-art primal federated optimization algorithms under certain situations. Numerical experiments on real-world datasets support our analysis.