Quantization-aware federated training with 8-bit parameter exchange cuts communication ~4x in polyp segmentation while keeping Dice within ~1.5 points of full precision.
Title resolution pending
1 Pith paper cite this work, alongside 364 external citations. Polarity classification is still indexing.
1
Pith paper citing it
364
external citations · OpenAlex
fields
cs.LG 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
QFedPolyp: A Communication- and Inference-Efficient Federated Learning Framework for Polyp Segmentation
Quantization-aware federated training with 8-bit parameter exchange cuts communication ~4x in polyp segmentation while keeping Dice within ~1.5 points of full precision.