Hierarchical Federated ADMM

TL;DR

This paper introduces a hierarchical federated learning framework based on ADMM, offering improved privacy, convergence, and accuracy over traditional gradient descent methods, with two novel algorithms for different layer configurations.

Contribution

The paper develops a new hierarchical federated learning framework using ADMM, including two algorithms that outperform conventional methods in privacy and performance.

Findings

ADMM-based hierarchical FL enhances privacy and accuracy.

Gradient descent performs well with limited local steps.

ADMM on both layers yields superior results.

Abstract

In this paper, we depart from the widely-used gradient descent-based hierarchical federated learning (FL) algorithms to develop a novel hierarchical FL framework based on the alternating direction method of multipliers (ADMM). Within this framework, we propose two novel FL algorithms, which both use ADMM in the top layer: one that employs ADMM in the lower layer and another that uses the conventional gradient descent-based approach. The proposed framework enhances privacy, and experiments demonstrate the superiority of the proposed algorithms compared to the conventional algorithms in terms of learning convergence and accuracy. Additionally, gradient descent on the lower layer performs well even if the number of local steps is very limited, while ADMM on both layers lead to better performance otherwise.