DFedRW runs parallel random walk model updates with decentralized averaging and reports accuracy gains of up to 38 percentage points over FedAvg and DFedAvg under high heterogeneity.
Adaptive random walk gradient descent for decentralized optimization,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.DC 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Decentralized Federated Averaging via Random Walk
DFedRW runs parallel random walk model updates with decentralized averaging and reports accuracy gains of up to 38 percentage points over FedAvg and DFedAvg under high heterogeneity.