Shadowheart SGD: Distributed Asynchronous SGD with Optimal Time Complexity Under Arbitrary Computation and Communication Heterogeneity

Abstract

We consider nonconvex stochastic optimization problems in the asynchronous centralized distributed setup where the communication times from workers to a server can not be ignored, and the computation and communication times are potentially different for all workers. Using an unbiassed compression technique, we develop a new method—Shadowheart SGD—that provably improves the time complexities of all previous centralized methods. Moreover, we show that the time complexity of Shadowheart SGD is optimal in the family of centralized methods with compressed communication. We also consider the bidirectional setup, where broadcasting from the server to the workers is non-negligible, and develop a corresponding method.

PDF NeurIPS OpenReview Semantic Scholar

Cite

Text

Tyurin et al. "Shadowheart SGD: Distributed Asynchronous SGD with Optimal Time Complexity Under Arbitrary Computation and Communication Heterogeneity." Neural Information Processing Systems, 2024. doi:10.52202/079017-0123

Markdown

[Tyurin et al. "Shadowheart SGD: Distributed Asynchronous SGD with Optimal Time Complexity Under Arbitrary Computation and Communication Heterogeneity." Neural Information Processing Systems, 2024.](https://mlanthology.org/neurips/2024/tyurin2024neurips-shadowheart/) doi:10.52202/079017-0123

BibTeX

@inproceedings{tyurin2024neurips-shadowheart,
  title     = {{Shadowheart SGD: Distributed Asynchronous SGD with Optimal Time Complexity Under Arbitrary Computation and Communication Heterogeneity}},
  author    = {Tyurin, Alexander and Pozzi, Marta and Ilin, Ivan and Richtárik, Peter},
  booktitle = {Neural Information Processing Systems},
  year      = {2024},
  doi       = {10.52202/079017-0123},
  url       = {https://mlanthology.org/neurips/2024/tyurin2024neurips-shadowheart/}
}