Convergence Analysis of Split Learning on Non-IID Data

Yipeng Li, Xinchen Lyu

Keywords: Federated Learning, Split Leanring, Convergence analysis

TL;DR: Convergence Analysis of Split Learning on Non-IID Data

Abstract: Split Learning (SL) is one promising variant of Federated Learning (FL), where the AI model is split and trained at the clients and the server collaboratively. By offloading the computation-intensive portions to the server, SL enables efficient model training on resource-constrained clients. Despite its booming applications, SL still lacks rigorous convergence analysis on non-IID data, which is critical for hyperparameter selection. In this paper, we first prove that SL exhibits an $\mathcal{O}(1/\sqrt{T})$ convergence rate for non-convex objectives on non-IID data, where $T$ is the number of total steps. By comparing the convergence analysis and experimental results, SL can outperform FL in terms of convergence rate (w.r.t. per-client training/communication rounds, and hence, the computation efficiency) and exhibit comparable accuracy to FL on mildly non-IID data. In contrast, FL prevails on highly non-IID data.

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.

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: Yes

Please Choose The Closest Area That Your Submission Falls Into: Deep Learning and representational learning