arXiv stat.ML
· Papers
Convergence of Sign-based Random Reshuffling Algorithms for Nonconvex Optimization
arXiv:2310.15976v4 Announce Type: replace-cross Abstract: signSGD is attractive in nonconvex optimization because it communicates sign-valued rather than full-precision gradients. Several standard analyses assume independent stochastic-gradient samples, whereas a common finite-sum implementation reshuffles the data and