Reactive Orchestration for Hierarchical Federated Learning Under a Communication Cost Budget

Ivan Čilić 1 , Anna Lackinger 3 , Pantelis A. Frangoudis 3 , Ivana Podnar Žarko 1 , Alireza Furutanpey 3 , Ilir Murturi 3 , Schahram Dustdar 3 · 2024

Read Reactive Orchestration for Hierarchical Federated Learning Under a Communication Cost Budget by Ivan Čilić 1 , Anna Lackinger 3 , Pantelis A. Frangoudis 3 , Ivana Podnar Žarko 1 , Alireza Furutanpey 3 , Ilir Murturi 3 , Schahram Dustdar 3 (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-03385.txt.

Description

Ivan Čilić and six co-authors present a framework that reorganizes federated learning on the fly in a 2024 paper. In hierarchical federated learning, phones and edge nodes train locally, passing updates through intermediate aggregators to a central server. Across the computing continuum, participants churn, costs bite, and data differ, so a fixed layout decays fast.

Their orchestration layer watches accuracy, resources, and prices at several levels, then triggers reshuffles when events demand it. A cost model for each reshuffle keeps decisions honest against the communication budget. Built on Kubernetes, the system reacted promptly in tests, trading cost and accuracy sensibly. The authors aim at learning pipelines that survive real-world volatility.

Catalog record

CREATORIvan Čilić 1 , Anna Lackinger 3 , Pantelis A. Frangoudis 3 , Ivana Podnar Žarko 1 , Alireza Furutanpey 3 , Ilir Murturi 3 , Schahram Dustdar 3
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE35 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-03385

Browse the full library · About the Tower of Babel