Derives a high-probability PAC-Bayesian bound on wireless generalization error for edge inference and proposes a channel-aware training algorithm minimizing a surrogate of the bound.
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
fields
cs.IT 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
A PAC-Bayesian Analysis of Channel-Induced Degradation in Edge Inference
Derives a high-probability PAC-Bayesian bound on wireless generalization error for edge inference and proposes a channel-aware training algorithm minimizing a surrogate of the bound.