Achieving Full Multiplexing and Unbounded Caching Gains with Bounded Feedback Resources - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2018

Achieving Full Multiplexing and Unbounded Caching Gains with Bounded Feedback Resources

Petros Elia

Résumé

In the context of the K-user MISO broadcast channel with cache-aided receivers, recent multi-antenna coded-caching techniques have sought to complement the traditional multiplexing gains associated to multiple (L) antennas, with the (potentially unbounded) caching gains (G) associated to coded caching. To date, all known existing efforts to combine the two gains, either resulted in a maximum known DoF L + G that required though CSIT on all (L + G) users served at a time (i.e., that induced potentially unbounded CSIT costs that matched the DoF gains), or resulted in a much compromised DoF where multiplexing gains came at the expense of bounded or vanishing caching gains. We present here a new multi-antenna coded caching algorithm that introduces a new XOR generation structure which completely untangles caching gains from CSIT, delivering the desired sum-DoF of L + G but with a much reduced CSIT cost of only L channel vectors at a time (L × L CSIT matrix). This means that for the first time in multi-antenna coded caching, one can achieve full multiplexing gains and unbounded caching gains, at the mere CSIT cost associated to achieving the multiplexing gains. In the end, the result solidifies the role of coded caching as a method for reducing feedback requirements in multi-antenna environments.
Fichier non déposé

Dates et versions

hal-03558150 , version 1 (04-02-2022)

Identifiants

Citer

Eleftherios Lampiris, Petros Elia. Achieving Full Multiplexing and Unbounded Caching Gains with Bounded Feedback Resources. ISIT 2018, IEEE International Symposium on Information Theory, Jun 2018, Vail, United States. pp.1440-1444, ⟨10.1109/ISIT.2018.8437704⟩. ⟨hal-03558150⟩

Collections

EURECOM ANR
27 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More