Incremental Aggregated Asynchronous SGD for Arbitrarily Heterogeneous Data
Keywords: distributed optimization, asynchronous SGD, data heterogeneity
Abstract: We consider the distributed learning problem with data dispersed across multiple workers under the orchestration of a central server. Asynchronous Stochastic Gradient Descent (SGD) has been widely explored in such a setting to reduce the synchronization overhead associated with parallelization. However, prior works have shown that the performance of asynchronous SGD algorithms depends on a bounded dissimilarity condition among the workers' local data, a condition that can drastically affect their efficiency when the workers' data are highly heterogeneous. To overcome this limitation, we introduce the Incremental Aggregated Asynchronous SGD (IA$^2$SGD) algorithm. With a server-side buffer, IA$^2$SGD makes full use of stale stochastic gradients from all workers to neutralize the adverse effects of data heterogeneity. In an asynchronous implementation setting, the algorithm entails two distinct time lags in the model parameters and data samples utilized in the server's iterations. Furthermore, by adopting an incremental aggregation strategy, IA$^2$SGD maintains a per-iteration computational cost that is on par with traditional asynchronous SGD algorithms. Our analysis demonstrates that IA$^2$SGD achieves a consistent convergence rate for smooth nonconvex problems for arbitrarily heterogeneous data. Numerical experiments indicate that IA$^2$SGD compares favorably with existing asynchronous and synchronous SGD-based algorithms.
Primary Area: optimization
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 5598
Loading