FedSub achieves O(rd) uplink communication and reduced gradient memory with a nonconvex convergence bound that includes a residual error floor from random projection variance.
Subspace optimization for large language models with convergence guarantees,
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.LG 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
support 1representative citing papers
citing papers explorer
-
An Efficient Subspace Algorithm for Federated Learning on Heterogeneous Data
FedSub achieves O(rd) uplink communication and reduced gradient memory with a nonconvex convergence bound that includes a residual error floor from random projection variance.