We propose a Minimum-Spanning-Tree-based scheduling and Multi-aggregation framework (MST-M) for communication-efficient Federated Learning. Simulation results show that MST-M saves over 10% in communication costs compared to existing heuristics.
Network for AI: Communication-Efficient Federated Learning with MST-based Scheduling and Multi-Aggregation over Optical Networks
Ibrahimi M.;Musumeci F.;Tornatore M.;
2024-01-01
Abstract
We propose a Minimum-Spanning-Tree-based scheduling and Multi-aggregation framework (MST-M) for communication-efficient Federated Learning. Simulation results show that MST-M saves over 10% in communication costs compared to existing heuristics.File in questo prodotto:
| File | Dimensione | Formato | |
|---|---|---|---|
|
OFC2024-ruikun.pdf
accesso aperto
:
Pre-Print (o Pre-Refereeing)
Dimensione
313.07 kB
Formato
Adobe PDF
|
313.07 kB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
