Ilyas Fatkhullin, Alexander Tyurin, Peter Richtárik
TL;DR本研究中我们提出的基于 Polyak 动量的 EF21-SGDM 算法在非凸随机优化中减少了之前 EF 算法中大批量的限制,同时后续提出的双动量版本进一步减少了计算复杂度。该算法不需要其他假设,并在理论证明上具有创新意义。
Abstract
Due to the high communication overhead when training machine learning models
in a distributed environment, modern algorithms invariably rely on lossy
communication compression. However, when untreated, the errors