Scalable and Resource-Efficient Second-Order Federated Learning via Over-the-Air Aggregation

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ghalkha, Abdulmomen, Issaid, Chaouki Ben, Bennis, Mehdi
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929675852316672
author Ghalkha, Abdulmomen
Issaid, Chaouki Ben
Bennis, Mehdi
author_facet Ghalkha, Abdulmomen
Issaid, Chaouki Ben
Bennis, Mehdi
contents Second-order federated learning (FL) algorithms offer faster convergence than their first-order counterparts by leveraging curvature information. However, they are hindered by high computational and storage costs, particularly for large-scale models. Furthermore, the communication overhead associated with large models and digital transmission exacerbates these challenges, causing communication bottlenecks. In this work, we propose a scalable second-order FL algorithm using a sparse Hessian estimate and leveraging over-the-air aggregation, making it feasible for larger models. Our simulation results demonstrate more than $67\%$ of communication resources and energy savings compared to other first and second-order baselines.
format Preprint
id arxiv_https___arxiv_org_abs_2410_07662
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Scalable and Resource-Efficient Second-Order Federated Learning via Over-the-Air Aggregation
Ghalkha, Abdulmomen
Issaid, Chaouki Ben
Bennis, Mehdi
Machine Learning
Second-order federated learning (FL) algorithms offer faster convergence than their first-order counterparts by leveraging curvature information. However, they are hindered by high computational and storage costs, particularly for large-scale models. Furthermore, the communication overhead associated with large models and digital transmission exacerbates these challenges, causing communication bottlenecks. In this work, we propose a scalable second-order FL algorithm using a sparse Hessian estimate and leveraging over-the-air aggregation, making it feasible for larger models. Our simulation results demonstrate more than $67\%$ of communication resources and energy savings compared to other first and second-order baselines.
title Scalable and Resource-Efficient Second-Order Federated Learning via Over-the-Air Aggregation
topic Machine Learning
url https://arxiv.org/abs/2410.07662