A learned rate-distortion codec trained jointly with a split-federated U-Net compresses transmitted features and gradients, achieving 1e3-1e4x data-transfer reduction at matched accuracy on skin-lesion segmentation, with smaller and inconsistent gains on blastocyst segmentation.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
SplitFedZip: Learned Compression for Data Transfer Reduction in Split-Federated Learning
A learned rate-distortion codec trained jointly with a split-federated U-Net compresses transmitted features and gradients, achieving 1e3-1e4x data-transfer reduction at matched accuracy on skin-lesion segmentation, with smaller and inconsistent gains on blastocyst segmentation.