AutoFLIP prunes federated models via one-time collective loss-landscape mapping and client-agreement-guided adaptation, reporting 52% lower computation and 65% lower communication with SOTA non-IID accuracy.
Federatedfactory: Generative one-shot learning for extremely non-iid distributed scenarios
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Pruning Federated Models through Loss Landscape Analysis and Client Agreement Scoring
AutoFLIP prunes federated models via one-time collective loss-landscape mapping and client-agreement-guided adaptation, reporting 52% lower computation and 65% lower communication with SOTA non-IID accuracy.