Optimal partitioning method among heterogeneous edge nodes in CNN models
Ryoya Okamoto, Xun Shao · 2024
In recent years, addressing the challenges of edge computing in executing complex CNN models has become crucial, with a focus on constructing lightweight models. While model compression can lead to accuracy degradation, model partitioning offers a method to reduce model size without compromising accuracy. This paper studies optimal partitioning methods between heterogeneous edge nodes, verifying the effectiveness through experiments.