Faster Adaptive Federated Learning

TL;DR

This paper introduces FAFED, an adaptive federated learning algorithm that improves convergence speed and efficiency using momentum-based variance reduction, addressing challenges of large models, datasets, and lack of adaptivity.

Contribution

The paper presents FAFED, the first adaptive federated learning algorithm with proven convergence guarantees and optimal sample complexity, enhancing efficiency over existing methods.

Findings

FAFED achieves optimal sample complexity of O(ε^{-3}) and O(ε^{-2}) communication rounds.

Experimental results show FAFED's superior efficiency on language modeling and image classification tasks.

Theoretical analysis confirms convergence without large batch requirements.

Abstract

Federated learning has attracted increasing attention with the emergence of distributed data. While extensive federated learning algorithms have been proposed for the non-convex distributed problem, federated learning in practice still faces numerous challenges, such as the large training iterations to converge since the sizes of models and datasets keep increasing, and the lack of adaptivity by SGD-based model updates. Meanwhile, the study of adaptive methods in federated learning is scarce and existing works either lack a complete theoretical convergence guarantee or have slow sample complexity. In this paper, we propose an efficient adaptive algorithm (i.e., FAFED) based on the momentum-based variance-reduced technique in cross-silo FL. We first explore how to design the adaptive algorithm in the FL setting. By providing a counter-example, we prove that a simple combination of FL and adaptive methods could lead to divergence. More importantly, we provide a convergence analysis for our method and prove that our algorithm is the first adaptive FL algorithm to reach the best-known samples $O(\epsilon^{-3})$ and $O(\epsilon^{-2})$ communication rounds to find an $\epsilon$-stationary point without large batches. The experimental results on the language modeling task and image classification task with heterogeneous data demonstrate the efficiency of our algorithms.