Prof. Kim’s paper “LocFedMix-SL: Localized, Federate, and Mix for Improved Scalability, Convergence, and Latency in Split Learning” has been accepted to the Web Conference 2022 (CS Conference Rank #8, H5-index=80, Impact Score=14.69, Acceptance Rate=17.7%).
This paper proposes a novel parallel split learning framework coined LocFedMix-SL to solve server-client update imbalance problem in split learning architecture.
This work was led by Seungeun Oh, Sihun Baek, Prof. Seong-Lyun Kim (above Yonsei University), Prof. Jihong Park (Deakin University), and Praneeth Vepacomma, Prof. Ramesh Raskar (above MIT).