Per-sample clipped SGD achieves optimal in-expectation and high-probability convergence rates for non-convex optimization under heavy-tailed gradient noise while outperforming standard SGD and batch clipping on CIFAR-100.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
math.OC 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Robust and Fast Training via Per-Sample Clipping
Per-sample clipped SGD achieves optimal in-expectation and high-probability convergence rates for non-convex optimization under heavy-tailed gradient noise while outperforming standard SGD and batch clipping on CIFAR-100.