Descrizione del progetto
Realizzare algoritmi di apprendimento automatico ecologici
L’ecocompatibilità delle nostre reti di comunicazione è una tendenza emergente, ma potenzialmente minacciata da algoritmi di intelligenza artificiale ad alta intensità energetica, necessari per complesse ottimizzazioni. Il progetto GreenML5G, finanziato dall’UE, studierà come ridurre il dispendio energetico di moduli di apprendimento di rinforzo profondo. Il risultato complessivo del progetto è la creazione di algoritmi di apprendimento automatico ecologici per la gestione delle risorse radio. L’impatto di questa tecnologia si diffonderà in altri settori dell’autonomizzazione e dell’apprendimento automatico.
Obiettivo
Artificial Intelligence (AI) is revolutionising a wide range of industries. Wireless networks with emerging high dimensional challenges are set to benefit from data-driven deep learning optimisation across layers. In particular, we expect that the deep supervised and deep reinforcement learning modules can resolve high-dimensionality inputs, achieve near optimal solutions, and efficiently scale via confederated learning. However, what is not well understood is the energy cost and carbon footprint of AI in future wireless networks. The danger is that intelligent networks are not green networks and that the recent progress made in green communication risk being undermined by the new breed of AI-based wireless communication. Here, in this project, we propose to develop green machine learning algorithms for radio resource management. This will lead to a future of intelligent and sustainable wireless networking.
Campo scientifico
- engineering and technologyelectrical engineering, electronic engineering, information engineeringinformation engineeringtelecommunicationstelecommunications networksmobile network5G
- natural sciencescomputer and information sciencesartificial intelligencemachine learningreinforcement learning
- natural sciencescomputer and information sciencesartificial intelligencemachine learningdeep learning
- engineering and technologyelectrical engineering, electronic engineering, information engineeringinformation engineeringtelecommunicationsradio technology
Parole chiave
Programma(i)
Argomento(i)
Meccanismo di finanziamento
MSCA-IF - Marie Skłodowska-Curie Individual Fellowships (IF)Coordinatore
MK43 0AL Cranfield - Bedfordshire
Regno Unito