Aggregate computing is a macro-approach for programming collective intelligence and self-organisation in distributed systems. In this paradigm, a single 'aggregate program' drives the collective behaviour of the system, provided that the agents follow an execution protocol consisting of asynchronous sense-compute-act rounds. For actual execution, a proper aggregate computing middleware or platform has to be deployed across the nodes of the target distributed system, to support the services needed for the execution of applications. Overall, the engineering of aggregate computing applications is a rich activity that spans multiple concerns including designing the aggregate program, developing reusable algorithms, detailing the execution model, and choosing a deployment based on available infrastructure. Traditionally, these activities have been carried out through ad-hoc designs and implementations tailored to specific contexts and goals. To overcome the complexity and cost of manually tailoring or fixing algorithms, execution details, and deployments, we propose to use machine learning techniques, to automatically create policies for applications and their management. To support such a goal, we detail a rich research roadmap, showing opportunities and challenges of integrating aggregate computing and learning.
Aguzzi G., Casadei R., Viroli M. (2022). Machine Learning for Aggregate Computing: a Research Roadmap. New York : Institute of Electrical and Electronics Engineers Inc. [10.1109/ICDCSW56584.2022.00032].
Machine Learning for Aggregate Computing: a Research Roadmap
Aguzzi G.;Casadei R.;Viroli M.
2022
Abstract
Aggregate computing is a macro-approach for programming collective intelligence and self-organisation in distributed systems. In this paradigm, a single 'aggregate program' drives the collective behaviour of the system, provided that the agents follow an execution protocol consisting of asynchronous sense-compute-act rounds. For actual execution, a proper aggregate computing middleware or platform has to be deployed across the nodes of the target distributed system, to support the services needed for the execution of applications. Overall, the engineering of aggregate computing applications is a rich activity that spans multiple concerns including designing the aggregate program, developing reusable algorithms, detailing the execution model, and choosing a deployment based on available infrastructure. Traditionally, these activities have been carried out through ad-hoc designs and implementations tailored to specific contexts and goals. To overcome the complexity and cost of manually tailoring or fixing algorithms, execution details, and deployments, we propose to use machine learning techniques, to automatically create policies for applications and their management. To support such a goal, we detail a rich research roadmap, showing opportunities and challenges of integrating aggregate computing and learning.File | Dimensione | Formato | |
---|---|---|---|
paper22-discoli-ac-rl-roadmap.pdf
accesso aperto
Tipo:
Postprint
Licenza:
Licenza per accesso libero gratuito
Dimensione
472.58 kB
Formato
Adobe PDF
|
472.58 kB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.