Derives a closed-form optimal gradient coding structure and bit allocation strategy to minimize residual error under an unbiasedness constraint for communication-efficient distributed learning in heterogeneous systems.
Performance guarantees of forward and reverse greedy algorithms for minimizing nonsu permodular nonsubmodular functions on a matroid,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
eess.SY 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Communication-Efficient Approximate Gradient Coding for Distributed Learning in Heterogeneous Systems
Derives a closed-form optimal gradient coding structure and bit allocation strategy to minimize residual error under an unbiasedness constraint for communication-efficient distributed learning in heterogeneous systems.