How WiFo-2 learns across heterogeneous wireless systems (IMAGE)
Caption
WiFo-2 is pretrained on LH-CSI, a heterogeneous three-dimensional space-time-frequency channel dataset, using masked denoising and a task-aware sparse mixture-of-experts architecture. The learned representation supports zero-shot channel reconstruction and few-shot adaptation across scenarios, system configurations, and communication and sensing tasks.
Credit
Boxun Liu et al./Science China Press
Usage Restrictions
Use with credit.
License
Original content