#machine learning
Nov 2025
Row-stochastic matrices can provably outperform doubly stochastic matrices in decentralized learning
A weighted Hilbert-space framework is developed and sufficient conditions under which the row-stochastic design converges faster even with a smaller spectral gap are derived, by using a Rayleigh-quotient and Loewner-order eigenvalue comparison.
Bing Liu, Boao Kong, Limin Lu et al.
· arXiv.org · 0 citations