A multilevel Monte Carlo framework debiases biased gradient compressors, preserving SGD convergence guarantees while reducing communication cost, with adaptive variance-minimizing level selection.
(2023), nonconvex case): 1 T TX t=1 E ∥∇f (xt)∥2 ∈ O ∆1L αT + ∆1Lσ1/2 α1/2T 3/4 + ∆1Lσ√ M T (101) Let us consider our bounds in Eq
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
-
Beyond Communication Overhead: A Multilevel Monte Carlo Approach for Mitigating Compression Bias in Distributed Learning
A multilevel Monte Carlo framework debiases biased gradient compressors, preserving SGD convergence guarantees while reducing communication cost, with adaptive variance-minimizing level selection.