@article{M94675496,
title = "Transmission Error Aware Split Layer Selection Scheme",
journal = "The Journal of Korean Institute of Communications and Information Sciences",
year = "2024",
issn = "1226-4717",
doi = "10.7840/kics.2024.49.6.839",
author = "Jaewook Lee, Keunsoo Ko, Haneul Ko",
keywords = "Split computing, transmission error, multi-armed bandit problem, AI service",
abstract = "Most of the works for improving split computing service performance did not consider the transmission error when the IoT device transmits the processing result of its sub-model (i.e., intermediated data) to the edge server. Unfortunately, the transmission error can degrade the inference result of the whole deep model due to the erroneous intermediated data. Thus, in this paper, we first analyze the effect of transmission error on the split computing service performance based on the simulation study. Then, we propose transmission error aware split layer selection (TESLS) to improve the service quality. In the TESLS scheme, the service controller interacts with the environment to find the best-split layer. The simulation results confirm that our proposed work is outperformed to the conventional split layer selection scheme."
}