ADSGD converges for non-convex decentralized optimization with computation-delay-independent step sizes and no bounded-heterogeneity assumption, via a reduction to asynchronous stochastic block coordinate descent.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Asynchronous Decentralized SGD under Non-Convexity: A Block-Coordinate Descent Framework
ADSGD converges for non-convex decentralized optimization with computation-delay-independent step sizes and no bounded-heterogeneity assumption, via a reduction to asynchronous stochastic block coordinate descent.