Ampere trains the device and server halves of a split model sequentially with a single activation transfer and consolidated server activations, cutting communication and computation while improving accuracy on non-IID data.
BottleNet++: An End-to-End Approach for Feature Compression in Device-Edge Co-Inference Systems,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.DC 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
background 1representative citing papers
citing papers explorer
-
Ampere: Communication-Efficient and High-Accuracy Split Federated Learning
Ampere trains the device and server halves of a split model sequentially with a single activation transfer and consolidated server activations, cutting communication and computation while improving accuracy on non-IID data.