Intelligent vehicles are quickly becoming mobile, powerful computers, able to collect, exchange; and process sensed data. They are therefore expected not just to consume ITS services, but also to actively contribute to the implemen- tation of relevant ITS applications. With an increasing role of machine learning (ML) approaches, vehicles are called to put into use their computing capabilities and sensed data for the training of ML models. This can be enacted through distributed learning approaches, which however may lead to significant communication overhead or to learners converging to different models. In this work, we envision a new distributed learning scheme, named EAGLE, that, with the assistance of the network edge, aims at exploiting the vehicles’ data and computing capabilities, while enabling an efficient learning process. To this end, EAGLE combines the advantages of two existing schemes, namely, federated learning and gossiping learning, yielding a distributed paradigm that ensures both scalability and model consistency. Our results, obtained using two different real-world data sets, show that EAGLE can improve learning accuracy by 20%, while reducing the communication overhead by 45%.
Edge-assisted Gossiping Learning: Leveraging V2V Communications between Connected Vehicles / Di Giacomo, G.; Haerri, Jerome; Chiasserini, Carla Fabiana. - STAMPA. - (2022), pp. 3920-3927. (Intervento presentato al convegno 25th IEEE Intelligent Transportation Systems Conference (ITSC 2022) tenutosi a Macau, China nel 08-12 October 2022) [10.1109/ITSC55140.2022.9921891].
Edge-assisted Gossiping Learning: Leveraging V2V Communications between Connected Vehicles
G. Di Giacomo;Carla Fabiana Chiasserini
2022
Abstract
Intelligent vehicles are quickly becoming mobile, powerful computers, able to collect, exchange; and process sensed data. They are therefore expected not just to consume ITS services, but also to actively contribute to the implemen- tation of relevant ITS applications. With an increasing role of machine learning (ML) approaches, vehicles are called to put into use their computing capabilities and sensed data for the training of ML models. This can be enacted through distributed learning approaches, which however may lead to significant communication overhead or to learners converging to different models. In this work, we envision a new distributed learning scheme, named EAGLE, that, with the assistance of the network edge, aims at exploiting the vehicles’ data and computing capabilities, while enabling an efficient learning process. To this end, EAGLE combines the advantages of two existing schemes, namely, federated learning and gossiping learning, yielding a distributed paradigm that ensures both scalability and model consistency. Our results, obtained using two different real-world data sets, show that EAGLE can improve learning accuracy by 20%, while reducing the communication overhead by 45%.File | Dimensione | Formato | |
---|---|---|---|
EAGLE_Paper.pdf
accesso aperto
Descrizione: Articolo principale
Tipologia:
2. Post-print / Author's Accepted Manuscript
Licenza:
PUBBLICO - Tutti i diritti riservati
Dimensione
620.49 kB
Formato
Adobe PDF
|
620.49 kB | Adobe PDF | Visualizza/Apri |
Chiasserini-Edge-Assisted.1.pdf
non disponibili
Tipologia:
2a Post-print versione editoriale / Version of Record
Licenza:
Non Pubblico - Accesso privato/ristretto
Dimensione
710.13 kB
Formato
Adobe PDF
|
710.13 kB | Adobe PDF | Visualizza/Apri Richiedi una copia |
Pubblicazioni consigliate
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/11583/2968099