Convergence bounds for decentralized gradient tracking using approximate finite-time consensus matrices show that approximation error continuously degrades consensus and steady-state error.
Distributed learning in non-convex en- vironments— part ii: Polynomial escape from saddle-points,
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
eess.SP 1years
2025 1verdicts
REJECT 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Decentralized Learning with Approximate Finite-Time Consensus
Convergence bounds for decentralized gradient tracking using approximate finite-time consensus matrices show that approximation error continuously degrades consensus and steady-state error.