arXivDaily arXiv每日学术速递 周一至周五更新
北美退伍军人节(Veterans Day),12 日无更新
arXiv 2602.03802cs.DCcs.AIcs.NAmath.NAmath.OC
收藏

我们需要异步SGD吗?关于同步方法的近优性的研究

Do We Need Asynchronous SGD? On the Near-Optimality of Synchronous Methods

Lomonosov Moscow State University(罗蒙诺索夫莫斯科国立大学)

Grigory Begunov, Alexander Tyurin

AI总结:

本文研究了同步SGD在异构计算场景中的近优性,证明其在多种情况下接近最优,尽管异步方法在某些情况下可能更合适。

摘要:

现代分布式优化方法大多依赖于传统的同步方法,尽管近年来在异步优化方面取得了显著进展。我们重新审视同步SGD及其稳健变体,称为m-同步SGD,并理论证明它们在许多异构计算场景中几乎是最优的,这有些出人意料。我们分析了在随机计算时间和对抗性部分参与工人的条件下同步方法,并证明了它们的时间复杂度在许多实际情况下是最佳的,最多相差对数因子。虽然同步方法不是万能的解决方案,存在某些任务可能需要异步方法,但我们证明它们足以满足许多现代异构计算场景的需求。

英文摘要:

Modern distributed optimization methods mostly rely on traditional synchronous approaches, despite substantial recent progress in asynchronous optimization. We revisit Synchronous SGD and its robust variant, called $m$-Synchronous SGD, which can be interpreted as a method with a backup-worker strategy, and theoretically show that they are nearly optimal in many heterogeneous computation and communication scenarios, which is somewhat unexpected. We analyze the synchronous methods under random times and adversarial partial participation of workers, and prove that their time complexities are optimal in many practical regimes, up to logarithmic factors. While synchronous methods are not universal solutions and there exist tasks where asynchronous methods may be necessary, we show theoretically that they are sufficient for many modern heterogeneous computation and communication scenarios.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑